decomposition, ignition and flame spread on furnishing ... · decomposition, ignition and flame...
TRANSCRIPT
Decomposition, Ignition and Flame Spread
on Furnishing Materials
A thesis submitted in fulfillment of the requirement
for the degree of
Doctor of Philosophy
by
Yun Jiang
Centre for Environment Safety and Risk Engineering
Victoria University, Australia
2006
ii
The general aim of this research is to find an effective and applicable method for prediction
of pyrolysis and ignition of certain furnishing materials in a real fire environment. Current
methods for combustion behaviour investigation use standardized bench-scale experiments,
such as cone calorimeter, Single Burning Item (SBI) and Lateral Ignition and Flame
Transport (LIFT) tests, to extensively and systematically collect data on the ignitability of
materials. The test data are useful for classification and evaluation of various building
materials, and even for computational schemes. However there are still serious limitations in
applying such experimental data directly on to the modelling of ignition and flame spread
under real fire environments when complicated geometry and flow conditions are major
concern. Meanwhile, current commonly used criteria for determination of ignition of solid
fuels were found incapable or inaccurate for the ignition prediction in the complicated
environments.
In the current study, certain furnishing materials, timbers, polyurethane foams and fabrics,
were chosen for research purposes. Some basic thermal properties of these furnishing
materials, such as thermal kinetic parameters and characteristic temperatures, were
measured. These thermal properties are critical for a better understanding of thermal
decomposition and ignition processes. Then they were applied to heat transfer and thermal
analysis through computer modelling to describe the decomposition process. Series of
bench-scale tests were carried out to construct a physical platform for modelling and
provide test results for validating of the modelling. Through fire modelling, various criteria
for ignition were investigated and compared with the test results. Particularly, a gas phase
parameter, mixture fraction, was validated against the lean flammability limitation theory
and relationship between them was found. Finally, based on the comparison and analysis, a
set of robust and accurate criteria, based on the critical mixture fraction, was promoted for
ignition and flame spread prediction as well as for improvement of current pyrolysis and
ignition models.
Decomposition process of the solid materials was investigated by thermogravimetric
analysis (TGA) test. Through current TGA tests, two sets of kinetics were observed in low
iii
and high heating rate ranges. The kinetics obtained from low heating rate is considered close
to their intrinsic values. The kinetics obtained from the higher heating rate range is believed
to be much more useful and effective in describing the pyrolysis process of real fire
conditions. Meanwhile Agrawal’s shifting pattern for a linear relationship between heating
rate and reaction rate in the high heating rate range was also found for the studied materials.
This shifting pattern provides the possibility to apply the “effective” kinetics to further
pyrolysis analysis and ignition prediction in real fire conditions.
Atreya’s one-step global reaction pyrolysis model was used in the decomposition modelling.
A CFD model, Fire Dynamics Simulator (FDS), which is based on Atreya’s pyrolysis and
heat transfer model, was used to perform the numerical simulation. By applying the
“effective” kinetics obtained from previously specified higher heating rate range TGA tests
for timber materials, a set of parameters describing the pyrolysis process was then obtained
from the simulations. These parameters include surface temperature, surface mass flux, and
mixture fraction. The simulated temperatures and mass fluxes match with the data from the
bench-scale cone calorimeter tests reasonably well.
Based on the lean flammability theory a critical mixture fraction concept was developed for
prediction of ignition phenomenon. Comparing with other commonly adopted ignition
criteria, such as critical surface temperature and critical mass flux, the critical mixture
fraction approach has the distinct advantage of being able to represent the mixture of
pyrolysis product and an oxidant, and the gas phase distribution of combustibles. By
carrying out numerical computation via FDS, distribution of the mixture fraction was
obtained. A relationship between this critical mixture fraction (at the ignition time
determined by previous cone calorimeter tests) and the lower flammable limit (LFL) has
been observed. A computation and prediction scheme for ignition of solid fuels was then
developed.
Through the study, a practical engineering approach to apply kinetics in the description of
decomposition and ignition has been built up. By adopting the critical mixture fraction, the
developed computation and prediction scheme is able extended to wider materials and
complicated fire environments to achieve a more accurate prediction without extensive
bench-scale tests.
iv
"!$# "!$# "!$# "!$#
The author wishes to acknowledge:
The endless support, encouragement and patience from his family.
The supervision of Professor Ian Thomas and Dr. Yaping He, and also the
assistance from all staff at the Centre for Environment Safety and Risk
Engineering, Victoria University of Technology.
Dr Justin Leonard, Division of Building, Construction and Engineering, CSIRO,
for his instruction and assistance on operation of cone calorimeter tests.
Professor Robert Shanks in Applied Chemistry, RMIT University, for providing
access to a Thermogravimetric analyser and guidance on the experiments.
Kevin McGrattan and Simo Hostikka, authors of the Fire Dynamics Simulator,
for their instruction on simulation and unpublished key functions of the model.
v
%&('*)+,+.-/1032%&('*)+,+.-/1032%&('*)+,+.-/1032%&('*)+,+.-/1032
I, Yun Jiang, declare that the PhD thesis entitled Decomposition, Ignition and Flame
Spread on Furnishing Materials is no more than 100,000 words in length, exclusive of
tables, figures, appendices, references and footnotes. This thesis contains no material that
has been submitted previously, in whole or in part, for the award of any other academic
degree or diploma. Except where otherwise indicated, this thesis is my own work.
Signature Date
vi
TABLE OF CONTENTS ABSTRACT ii
ACKNOWLEDGEMENTS iv
DECLARATION v
NOTATION xi
LIST OF ABBREVIATIONS xv
Chapter 1 INTRODUCTION 1
1.1 Background 1
1.2 Research Requirements for Ignition and Flame Spread Phenomena of Solid Materials 3
1.2.1 Thermal decomposition and kinetics 3 1.2.2 Combustion models and numerical solutions 4 1.2.3 Furnishing materials 6
1.3 Aims of Research 7
1.4 Significance of Research 7
1.5 Research Methodology 8
1.5.1 General methodology 8 1.5.2 Methods involved in key activities 8
1.6 Brief Overview of this Thesis 9
Chapter 2 LITERATURE REVIEW 11
2.1 Background 11
2.2 Fire Models 15
2.2.1 Zone models 15 2.2.2 Field models 18
2.2.2.1 Computational Fluid Dynamics (CFD) basics 18 2.2.2.2 Turbulence 20
2.2.3 Comparison between the zone and field models 21 2.3 Thermal Decomposition 22
2.3.1 Physical and chemical processes 23 2.3.2 Kinetics 25 2.3.3 Decomposition of the furnishing materials 27
2.3.3.1 Pyrolysis and charring of wood 28 2.3.3.2 Role of moisture content 30
2.4 Ignition Controlling Mechanism 31
vii
2.4.1 Auto Ignition 32 2.4.1.1 Solid heating and gasification 34 2.4.1.2 Gas phase reaction 38 2.4.1.3 Ignition time, tig 40 2.4.1.4 High surface temperature ignition 40
2.4.2 Piloted ignition 41 2.4.3 Ignition criteria 42
2.4.3.1 Critical surface temperature 43 2.4.3.2 Critical mass flux 45 2.4.3.3 Lower flammable limit and possible criteria in gas phase 46
2.5 Flame Spread on Solid Combustible Surfaces 50
2.5.1 Flame spread mechanism 50 2.5.2 Material properties and measurement for the flame spread 53
2.6 Experimental Methods 53
2.6.1 Thermogravimetric analysis (TGA) 54 2.6.2 Cone calorimeter 56 2.6.3 LIFT and SBI 57 2.6.4 Applications of the experimental methods 58
2.7 Unresolved Phenomena and Research Interests 60
2.7.1 Unresolved issues related to current research 60 2.7.2 Research interests and requirements identified for current study 62
Chapter 3 MODELLING OF THERMAL DECOMPOSITION AND IGNITION OF SOLID MATERIALS 64
3.1 Pyrolysis and Ignition Models 64
3.1.1 Classification of pyrolysis models 64 3.1.2 Typical pyrolysis models 66 3.1.3 Criteria of model selection for current research 68
3.2 Atreya’s One-dimensional Heat Transfer and Pyrolysis Model 70
3.2.1 Basics of Atreya’s pyrolysis model 70 3.2.1.1 Physical model 70 3.2.1.2 Four stages of wood pyrolysis 71
3.2.2 Basic equations and assumptions 73
3.3 Improvements and Applications for Atreya’s Model 75
3.3.1 Improvement by Parker 75 3.3.2 Improvement by Ritchie et al. 76
3.4 Calculation for a Studied Material Based on Atreya’s Model 78
3.4.1 Thermal properties of studied material 78 3.4.2 Calculation results from Atreya’s model 79
3.5 Summary 80
viii
Chapter 4 EXPERIMENTAL DECOMPOSITION STUDY FOR THE FURNISHING MATERIALS 81
4.1 TGA Experiment and its Application 81
4.1.1 TGA apparatus, principle and test procedure 81 4.1.2 Research methodology 83
4.1.2.1 Curve trend and characteristic temperatures 83 4.1.2.2 Kinetics and calculation method 84 4.1.2.3 Concern for test conditions and computation schemes 88 4.1.2.4 Relationships among the parameters 90
4.2 Current TGA Test Results 92
4.2.1 Experimental materials and apparatus 92 4.2.1.1 Details of materials 92 4.2.1.2 Apparatus and test conditions 93 4.2.1.3 Calibration 93
4.2.2 Measurement of thermal characteristics of the materials 93 4.2.2.1 Decomposition process 94 4.2.2.2 Effect of heating rate 97
4.2.3 Thermal dynamic parameters and behaviours 101 4.2.3.1 Activation energy and pre-exponential factor 101 4.2.3.2 Linear relationships and shifting pattern 104
4.2.4 Isothermal test 106
4.3 Summary 108
Chapter 5 CONE CALORIMETER STUDY FOR IGNITION PROPERTIES 110
5.1 Heat Release Rate Measurement 111
5.1.1 Oxygen consumption method 111 5.1.2 Cone calorimeter 112
5.2 Calibration and Operation of Cone Calorimeter 115
5.2.1 Calibration of cone calorimeter 115 5.2.2 Experiment procedure 117
5.2.2.1 Testing materials and preparation 117 5.2.2.2 Testing condition and data sampling 119
5.3 Experimental Results 120
5.3.1 Ignition time 120 5.3.1.1 Tested results 120 5.3.1.2 Linear relationship between 5.0−
igt vs radiation 121 5.3.1.3 Determining the critical radiative flux 123 5.3.1.4 Effect of sample thickness 124 5.3.1.5 Comparison with results from others 125
5.3.2 Heat release 125 5.3.2.1 Heat release rate 126 5.3.2.2 Total heat released 130
ix
5.3.3 Mass loss rate and effective heat of combustion 130 5.3.4 Species yield 134
5.4 Surface Temperature Measurement 139
5.4.1 Surface temperature 139 5.4.2 Temperature rising rate 144
5.5 Summary of the Cone Calorimeter Test Results 145
Chapter 6 SIMULATING PYROLYSIS AND IGNITION BY FDS FOR THE WOODS 148
6.1 FDS Modelling for Pyrolysis and Combustion 148
6.1.1 Principle of FDS model 149 6.1.1.1 History of major developments 150 6.1.1.2 Mixture fraction combustion model 151 6.1.1.3 Pyrolysis and charring models in FDS 153 6.1.1.4 Ignition criteria in FDS 154
6.1.2 Treatment of thermal parameters for timber in FDS simulation 155 6.1.2.1 Mixture fraction relation for timber 155 6.1.2.2 Calculation for the thermal dynamic parameters 157
6.2 Setting Up and Running of FDS Simulations 159
6.2.1 General setting 159 6.2.1.1 Simulation domain and grid size 159 6.2.1.2 “Cone” heater 161
6.2.2 Heater temperature and heat flux 161 6.2.3 Cell grid 164 6.2.4 Summary of input parameters in FDS simulation 168
6.3 Simulation Results 169
6.3.1 Observation of initial flame development 169 6.3.2 Surface temperature 174 6.3.3 Heat release rate 175 6.3.4 Mass flux 177 6.3.5 Direct Numerical Simulation 181 6.3.6 Comparison on results from different E and A values 182 6.3.7 Effect of sample thickness and backing material 183
6.4 Distribution of Pyrolysis Products 186
6.5 Summary of the Simulation Results 188
Chapter 7 CRITERIA FOR IGNITION AND DISCUSSION 191
7.1 Further Discussion of Ignition in FDS 191
7.2 Comparison Between the Simulation and Experimental Data 194
7.2.1 Critical ignition temperature 195 7.2.2 Critical mass flux 195
x
7.2.3 Critical mixture fraction 196
7.3 Special Consideration for the Critical Mixture Fraction 197
7.3.1 Relationship between the critical mixture fraction and the LFL 197 7.3.1.1 Calculation of LFL for timber 197 7.3.1.2 Converting the mixture fraction into volume fraction 198
7.3.2 Discussion for the suggested criterion 200
Chapter 8 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH 203
8.1 Conclusion 203
8.2 Further Research Recommendations 204
APPENDIX A. TGA TEST SETTINGS AND RESULTS 206
APPENDIX B. CONE CALORIMETER TEST SETTINGS AND RESULTS 209
APPENDIX C. ANALYSIS AND COMPUTATION OF ATREYA’S PYROLYSIS MODEL 216
REFERENCES 223
xi
ê ëEì\í.ìî1ë3êê ëEì\í.ìî1ë3êê ëEì\í.ìî1ë3êê ëEì\í.ìî1ë3ê
A pre-exponential factor in Arrhenius equation (1/s)
[A] concentration of the reactant A (mol/m3)
B body force in momentum conservation equation (N)
c specific heat (J/kg⋅K)
cp isobaric specific heat (J/kg⋅K)
cv isochoric specific heat (J/kg⋅K)
C coefficient for linearized radiative heat transfer (kW/m2)
Ci volume percent of fuel, i, in fuel/air mixture
Dp degree of polymerization, number of monomer units per polymer chain
e base of natural logarithms
E activation energy (kJ/mol)
EHPM heat release per mass unit of oxygen consumed (kJ/kg)
F external heat flux (kW/m2)
h static (thermodynamic) enthalpy (J/kg)
h heat transfer coefficient (kW/m2⋅K)
H total enthalpy (J/kg)
H reaction rate (1/s)
Hremove rate of heat removed from solid surface (W/m2·K)
∆Hc heat of combustion (kJ/kg)
k turbulent kinetic energy in k-ε equation
k reaction rate constant (1/s)
L slab thickness (m)
L effective heat of gasification (kJ/kg)
Lphy physical length in thermal thickness judgment (m)
Lthermal thermal length characteristic in thermal thickness judgment (m)
Lv heat of vaporization (kJ/kg)
M mass of material (kg)
MWi molecular weight for species i (g/mol)
m mass flow rate (kg/s)
m ′′ mass flux (kg/m2⋅s)
n reaction order
Fp pressure fraction of fuel
Notation ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
xii
"q heat flux per unit area (W/m2)
"radq external radiation flux (W/m2)
"rsq re-radiation flux from the solid surface (W/m2)
Q ′′′ critical energy density (kJ/m3)
r radius (m)
R universal gas constant, = 8.314×10-3 (kJ/mol⋅K)
Ri,cr critical values in ignition criteria for item i, (i =1, 2, …)
s solid fuel thickness (m)
s ratio of consumption rates for the fuel and oxygen in FDS
S section area (m2)
S source or sink in mass conservation equation
t time (s)
tin gas induction time (s)
T temperature (K)
Tf,LFL adiabatic flame temperature of a lower flammable limit mixture (K)
u velocity parallel to the solid surface (m/s)
u∞ forced flow free stream velocity (m/s)
U fluid velocity vector (m/s)
v velocity (m/s)
vi stoichiometric coefficient for species i
W sample weight (kg)
W∞ final sample weight (kg)
x fraction of conversion of sample weight
x coordinate parallel to the solid surface
y coordinate normal to the solid surface
Y mass fraction
Z mixture fraction (kg/kg)
Z-surface: mixture fraction at surface of solid;
Z-plug: mixture fraction at the spark plug in cone calorimeter
Greek Symbols
α thermal diffusivity (m2/s)
Notation ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
xiii
β non-dimensional parameter in gas phase ignition analysis,
β=T∞ cp/YF,s∆Hc
β linear heating rate (K/s)
Γ non-dimensional parameter in gas phase ignition analysis,
Γ=4c(E/RT∞)[(2-β)/(e2(1-β2))]
δ boundary layer thickness (m)
ε local energy in k-ε equation
thermal conductivity (W/m⋅K)
φ combustion efficiency
φ item in measurement of flame heat transfer item
Λo characteristic Damkohler number in the stagnation point flow ignition analysis
µ molecular viscosity (kg/m⋅s)
ν kinematic viscosity (m2/s)
ρ density (kg/m3)
σ item in momentum conservation equation
σ Stefan-Boltzmann radiation constant (W/m2⋅K)
σ non-dimensional parameter in gas phase ignition analysis,
σ=Λo(1-exp(-4u∞t/y))/(cE/RT∞)
Subscripts
a active
air air
c char
cr critical
eff effective
ev evaporation
f flame
F fuel
g gas
I initial
ig ignition
Notation ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
xiv
Inert inert
m moisture
max maximum
min minimum
oxy oxygen
pd product
py pyrolysis
s surface
samp sample
solid solid phase
spec species
∞ Ambient or final
w wood
Superscripts
* Non-dimensional parameter
average
″ flux
flow rate
xv
ïð_ñòôóõö÷R÷øùJú\ðûö.òð1óüdñïð_ñòôóõö÷R÷øùJú\ðûö.òð1óüdñïð_ñòôóõö÷R÷øùJú\ðûö.òð1óüdñïð_ñòôóõö÷R÷øùJú\ðûö.òð1óüdñ
ASTM American Society for Testing and Materials
CFD computational fluid dynamics
DNS direct numerical simulation
DTG differential thermogravimetric
FDS Fire Dynamics Simulator, a fire model developed at NIST
FIST Forced-flow Ignition and flame-Spread Test
FSE Fire Safety Engineering
FSP Fibre Saturation Point
FVM Finite volume method
HRR heat release rate
HRRPUV heat release rate per unit volume
LES Large Eddy Simulation
LFL lower flammable limit
LIFT Lateral Ignition and Flame Transport
LOI lower oxygen index
MF mixture fraction
MLR mass loss rate
MW molecular weight
N-S Navier-Stokes (equation)
ODE ordinary differential equations
OFW Ozawa-Flynn-Wall method
PDE partial differential equations
PMMA polymethylmethacrylate
PU Polyurethane foam
SBI Single Burning Item
TG thermogravimetric
TGA Thermogravimetric analysis
THR total heat released
VF volume fraction
1
ý*þRÿý*þRÿý*þRÿý*þRÿ ý ý ý ý
1.1 Background
The key to widely and effectively applying performance-based building design and risk
assessment is to quantitatively predict fire behaviour and performance of buildings and
components. Performance-based building codes have been introduced into more and more
countries, including Australia, since the 1990s. Performance-based building design enables
engineers to evaluate the performance of buildings and components by considering a wide
range of realistic fires. This provides great flexibility for adopting new concepts, materials
and technologies in building designs, while simultaneously achieving lower costs and lower
risk to life. The performance-based approach is usually achieved by some type of
performance-oriented model as a design and assessment tool. CESARE-RISK is such an
example (Beck 1998). Among all sub-models in a performance-based building design and
risk assessment model, the combustion and fire growth model, which describes the fire
origin and development, is the cornerstone. It provides various outputs as critical inputs for
other sub-models to assess building damage and risk to life, through barrier failure model,
economic model, smoke spread model, and human behaviour and evacuation models.
Therefore, a better understanding of real fire development and building performance in fires
is a foundation for performance-based design and risk assessment tasks.
Today there are several standardized experimental methods to measure the performance of
materials and components in fire environments, as the basis for prediction of fire behaviour.
These experimental methods include:
bench-scale testing devices, such as the cone calorimeter (Babrauskas 1984) and
the Lateral Ignition and Flame Transport (LIFT) (Quintiere 1981)
middle-scale testing devices, like the Single Burning Item (SBI) (Mierlo 1998),
and the furniture calorimeter
full-scale methods, such as the room/corner burning test and room calorimeter
(ASTM 1983).
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
2
These methods have been used in many international projects, such as the SBI Round Robin
Tests (Mierlo 1997; VTT 1997) and the Cone Calorimeter Inter-laboratory Trials (ASTM
1990). These standardized experimental methods play an important role in establishing
databases for assessing material properties, material performance, and combustion
behaviour in fire environments.
However, due to the huge variety of building materials and their combinations as building
structures and components, it is impossible to test every material and structural
configuration in the various scale tests. There are also significant differences between
standardized testing conditions and fire environments that occur in buildings, due to the
variety and complexity of building geometries and ventilation conditions. Therefore
experimental methods cannot be the exhaustive means of investigation of fire performance.
Modelling is another way to study building performance under fire conditions. A fire model
is a statement in mathematical language to describe fire phenomena. Fire models now are
able to simulate several aspects of fire scenarios (including geometry, ventilation, fuel
configuration and even reaction of human beings) over an acceptable modelling duration.
Fire modelling is an effective and economical approach for predicting a fire outcome before
a fire happens and representing the history of fire development after it happens. During
recent decades, fire models have greatly benefited from rapidly developing computer
technology. Models available for fire engineers, which were once relatively primitive, are
now comprehensive, and can deal with quite complicated phenomena.
There are still many challenges for today’ s fire models to achieve an applicable and
acceptable performance for building fire scenarios (Cox 1994). Firstly the thermal
properties of the fuels and other contents or materials in an enclosure must be known for
inputting into the models. As indicated by Babrauskas (1996), “ the most serious limitation,
however, comes from the fact that essential items of fire physics and chemistry are missing
from even the best of the existing models.” Secondly, the fire models should be robust and
usable by fire protection engineers. A good fire model should deal with as many details as
possible within an acceptable duration of computation. Finally, prediction of real fires must
be validated by experimental results and on-site fire surveys (after a fire incident has
happened). Currently adopted ignition criteria are still questionable against on wider variety
of fuel materials and conditions. Bridging gaps between current fire models and their
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
3
application in fire safety engineering is a critical task for today’ s fire research and a major
aim of this study. To achieve this aim, quantitative analysis of basic aspects involving
combustion and development of building fires is essential.
1.2 Research Requirements for Ignition and Flame Spread Phenomena of
Solid Materials
As identified in the previous section, ignition and combustion phenomena are major
research interests for this study. These phenomena involve basic aspects of physics and
chemistry of fire. For example, a spreading flame includes many interactive sub-processes:
fluid flow, heat and mass transfer, solid thermal decomposition, gas-phase chemical kinetics,
etc. A comprehensive solution for modelling fire behaviours requires extensive knowledge
of the phenomena, as well as the application of detailed modelling.
1.2.1 Thermal decomposition and kinetics
Ignition of solid materials can be thought of as a twofold process:
the decomposition of the solid phase of fuel under an external heating source,
and
the release of volatile products and a gas phase oxidizing reaction.
These two processes are interlaced. When certain ignition criteria are reached, mainly in the
gas phase, ignition takes place.
There is a long history of the study of thermal decomposition phenomena of solid fuels,
starting with timber and coal. Beyler and Hirschler (2002) give a general and detailed
description for the decomposition process of polymers (including timber and coal). A
number of physical and chemical processes may be involved in thermal decomposition.
These include melting, charring, and vaporization, etc. Beyler and Hirschler listed eight
generic types of reaction involved in a simple decomposition process. They also indicated
that the kinetics describing these reactions could be quite complex and unusable in
engineering applications.
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
4
Therefore simple overall kinetic expressions are usually utilized as a reasonable engineering
approach. It should be pointed out that nearly every current fire model, which deals with
ignition and combustion of solid materials, involves simplification of one sort or another
(Moghtaderi 2001). The most common assumption in such simplifications is that the
reactions can be described by a one-step Arrhenius expression involving the remaining solid
mass. However, characterizing those reactions or even defining a simplified overall
decomposition process to represent those complicated reactions is still a challenge for
today’ s fire models.
Several experimental methods have been developed for the study of thermal decomposition
(Beyler and Hirschler 2002). Thermogravimetric analysis (TGA) is commonly used. The
kinetics obtained from TGA tests have been adopted to describe either an overall thermal
decomposition process (such as for a timber) or reactions by individual components (such as
cellulose, hemicellulose and lignin in a timber) (Baker 1978). This method has been applied
in fire safety science and engineering for many years. It is noted that there are still some
limitations in this method due to the difference between the test conditions and real fire
environments. One major issue is the lower heating rate normally adopted in the TGA tests.
The normally adopted heating rate in a TGA test is lower than 30 K/min (0.5 K/s) while the
rate of rising of surface temperature can reach a magnitude of 10 to 100 K/s in a real fire
scenario (Beyler and Hirschler 2002). Therefore how to apply kinetics from this type of
general-purpose chemical analysis test into description of thermal decomposition properly
in a real fire condition still remains as one of the major tasks in solid pyrolysis study.
1.2.2 Combustion models and numerical solutions
Combustion models are a group of sub-models in fire models, dealing with various aspects
of fire initiation and development. These sub-models simulate the pyrolysis of fuel,
charring, ignition, flame spread over a solid surface, pool fire, radiation heat transfer and
generation of combustion products. Based on modelling results from these sub-models, the
combustion model predicts fire development and outcome, and provides two kinds of major
information about a fire: heat release rate and product gases yield. This information is
important for assessing either building damage or life safety. Therefore, the combustion
model is a critical component within any fire model.
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
5
Depending on how many details are included, a combustion model can vary from a very
simple, such as a plume model found in some early fire models, to a quite complicated
comprehensive model which may include most of the above mentioned aspects of fire
physics and chemistry.
Fire models are normally categorised into two types: zone models and field models, and the
latter are also known as computational fluid dynamics (CFD) models. The zone models are
based on a conceptual representation for the compartment fire process and are good
approximations to reality for certain fires. When distinct phenomena are discerned and well
isolated, they may provide the ability for better predictions of their roles in the compartment
fire system (Quintiere 2002). By contrast, CFD models solve the governing equations by
numerical methods and become a general tool for the analysis of the full breadth of fluid
flow problems, including those associated with fire. The CFD type fire models are more
capable of studying complicated compartment fire problems due to their ability to deal with
fine grid details of phenomena, and provide much more accurate prediction for fire
development.
However, CFD fire models also have limitations. It is impossible to model all details
involved in combustion. First, due to a lack of better understanding of fire phenomena and
dynamics, simplification is always necessary. For example, a global reaction is usually
chosen to represent a complicated decomposition process. Secondly human expectations
from the model usually tend to exceed computational capabilities because of the complexity
of compartment fire. Cox and Kumar (2002) provide a simple example. To capture the
details of a chemical reaction zone in a fire would require a characteristic mesh size below 1
mm. Such a computational mesh would require computer power and time that are affordable
to only a few. In order to obtain solutions with reasonable accuracy and affordable
resources, it is necessary to simplify the system of equations by some form of
approximation.
Meanwhile for most current pyrolysis and ignition models, no matter what kind of
numerical computation schemes adopted, no widely applicable ignition criteria available is a
general and serious limitation. Currently adopted parameters, such as critical surface
temperature, critical temperature rising rate, and critical pyrolysis rate, have been found
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
6
questionable and great variations have been reported for test data. This limitation will be
reviewed in detail in next chapter.
To perform a detailed study on pyrolysis and ignition phenomena, including solid phase
heat transfer and gas phase distribution, the CFD model is a suitable tool for this research.
As mentioned before, choosing the details to be included or excluded is an important task in
developing and using fire models. Improving computational schemes to find more efficient
numerical solutions is another task. Finally, determining a set of applicable ignition criteria
to wide application areas of fire modelling is also a challenge.
1.2.3 Furnishing materials
Furnishing materials often play an important role in a compartment fire. They can be not
only the original fuel of fires, but also the means of flame spread in enclosures. Furnishing
materials cover a wide range of materials, from natural materials, such as timber and cotton
fabrics, to artificial polymers, such as PVC and polyurethane (PU) foam. Timber is often
widely used as for structural components as well as for furniture material. Polyurethane
foam is extensively used in mattresses and as cushions in sofas and chairs. Fabrics are a
common material for covering these objects and as curtains. All these types of materials
have been frequently related to fire initiation and flame spread in compartment fires (Ellis
1981; DeHaan 2002).
A great number of research projects have been undertaken on these materials, especially on
timbers. These researches will be reviewed in some detail in Chapter 2. However, due to the
variety and complexity of these materials, as well as the almost endless combinations of
them, there is still much uncertainty remaining in modelling fire behaviours for these
materials. In particular, knowledge of ignition criteria for many of these types of materials is
quite limited. Therefore, two of each of these materials were chosen for the current research.
Thermal decomposition and kinetics of these selected materials will be the key research
interest since they determine the ignition phenomenon.
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
7
1.3 Aims of Research
The general aim of this research is to develop a new method to quantitatively describe
pyrolysis, ignition and combustion processes for several furnishing materials.
By solving uncertainty and requirements that were identified in the previous section, this
general aim has been achieved via the following sub-aims:
1. Improve an existing experimental and computational scheme to obtain kinetic data
for the studied furnishing materials. The obtained kinetics are able to describe the
thermal decomposition behaviour under real fire conditions.
2. Collect thermal properties of the thermal decomposition and ignition processes for
the studied materials via various scaled experimental methods. These experimental
data were used either to provide input information for further modelling or to
validate the simulation results.
3. Identify an appropriate ignition criterion by applying previously obtained thermal
properties into a pyrolysis and ignition model via a suitable numerical solution (CFD
model).
1.4 Significance of Research
The significance of this research includes the following:
Different decomposition behaviours under various environmental conditions have
been observed for the studied furnishing materials. The conditions and limitations to
apply a simplified overall decomposition model are identified. These contribute to a
better understanding of the basic thermal decomposition of the materials.
Applicable thermal kinetics have been obtained for the materials to describe the
pyrolysis process in realistic fire conditions. These calculated parameters will enable
more accurate modelling results to be generated. This has demonstrated an effective
way to apply small-scale, basic thermal testing results into simulation of bench-scale
to full-scale tests.
Ignition and combustion properties under bench-scale tests have been obtained.
These results have added to the knowledge database of the studied furnishing
materials. The experimental data can be used to validate modelling results.
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
8
A more widely applicable gas phase ignition criterion, critical mixture fraction, for
the studied materials has been found which is satisfactory for engineering
calculations. The criterion was found related to lean flammability limit theory. This
provides the possibility to improve the current pyrolysis and ignition models and
extend the application area of the models to more complicated environments
accurately.
1.5 Research Methodology
1.5.1 General methodology
The following sequence has been adopted in this thesis, reflecting the generally accepted
research methodology:
1. Literature review
2. Review and further development of theoretical modelling of pyrolysis and ignition
processes for solid materials
3. Experimental studies of the thermal decomposition and ignition processes of the
chosen furnishing materials. These tests include basic kinetic experiments and
bench-scale tests for combustion behaviour
4. Computer modelling to simulate decomposition and ignition processes in the test
environment
5. Analysis and discussion of ignition criteria leading to the proposal of a more
applicable criterion
1.5.2 Methods involved in key activities
The methods adopted for the kinetics calculations are:
Review and evaluate current experimental methods and calculation schemes for the
decomposition and chemical kinetics
Find some pattern or relationship to enable those parameters, which are used to
describe an overall decomposition process, to be applicable in an environment closer
to real fire conditions
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
9
Determine the “ effective” kinetics for the materials by the identified method
The methods involved in the collection of properties of pyrolysis and ignition processes are:
Carry out bench-scale fire tests of the materials
Measure ignition and combustion properties of the materials, especially for those
related to ignition criteria
Identify the factors affecting the ignition process
The methods involved in the evaluation of the pyrolysis model and its numerical solution
are:
Review available mathematical fire models dealing with ignition and combustion.
The basic requirements will include a kinetic description for pyrolysis of charring
materials, such as timber and fabrics
Use a CFD model to perform the numerical solution for the mathematical model
Run extensive simulations of the studied materials to obtain ignition properties, and
the discrete relationship between those parameters
The methods involved in ignition criteria study are:
Link and compare the experimental and modelled ignition properties for the studied
materials
Evaluate various ignition criteria for the materials against the experimental data and
lean flammability theory
Find one or several ignition criteria appropriate for the tested materials and compare
and validate with the experimental data and research results obtained by others
1.6 Brief Overview of this Thesis
In line with this approach, this thesis has been organized in the following manner:
Chapter 2 is a general review of areas related to the current research. These include: fire
models, thermal decomposition, ignition and criteria, flame spread mechanisms of solids,
experimental methods, and thermal properties of the studied furnishing materials. Unsolved
problems are also identified to define research areas and tasks.
Chapter 1 Introduction ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
10
Modelling for pyrolysis and flame spread on a solid is discussed in depth in Chapter 3.
Various fire models involving ignition and combustion are evaluated and compared. Based
on comparison and analysis, an appropriate pyrolysis model is then chosen as a prototype
for the current research.
In Chapter 4, TGA experiments are carried out to determine the thermal properties of the
studied materials. Based on the experimental data, kinetics are calculated for the materials.
An application scheme to apply the “ effective” kinetics into an environment closer to real
fire conditions is then developed.
In Chapter 5, a set of bench-scale tests, cone calorimeter tests, are performed to obtain the
ignition and combustion behaviours for the same materials. Some other parameters, such as
surface temperature and ignition time that are helpful to validate the pyrolysis model, are
also measured from the tests.
In Chapter 6, a CFD fire model is used to simulate the decomposition and ignition processes
in a confined environment, simulating the cone calorimeter test. The thermal properties,
including the kinetics obtained from the previous TGA tests, are used as input parameters.
The simulation results are compared with and then validated by the cone calorimeter test
data.
In Chapter 7, several available ignition criteria are compared with each other as well as the
experimental data. Through analysis and discussion, a robust ignition criterion is then
proposed.
A final summary and conclusion is given in Chapter 8.
11
! ! ! ! "$#%&')(*"$#%&')(*"$#%&')(*"$#%&')(*+-,.#0/-,.#0/-,.#0/-,.#0/
This review consists of seven sections. The first section briefly introduces the background
of the current research. As indicated in the first chapter the research will focus on solid fuel
ignition and combustion phenomena. Section 2.2 outlines computer fire models used in fire
engineering. Two main kinds of computer fire models of fire in enclosures (zone and field
models) are introduced briefly and compared. Section 2.3 reviews previous major research
activities on thermal decomposition of solid materials and various schemes for determining
the kinetics of decomposition. A special discussion of the decomposition process of timber
is given in this section. The controlling mechanism of ignition of solid fuel is then described
in Section 2.4. Major factors related to ignition as well as the criteria of ignition are
reviewed. Section 2.5 describes the mechanism of flame spread on solid fuel materials.
Experimental methods and their application for obtaining thermal properties and
combustion behaviour are discussed in Section 2.6. Unresolved issues in fire modelling,
pyrolysis and ignition processes are identified in Section 2.7. Based on this discussion,
current research interests and requirements are then determined.
2.1 Background
During the past two decades, performance-based building codes and design practices have
begun to be applied in many countries. In Australia, the first performance-based building
code was released nearly ten years ago (ABCB 1996). More and more building projects are
adopting the performance-based approach and “ alternative solutions” are used rather than
the deemed-to-satisfy solutions included in the Building Code Australia (BCA) and previous
documents. Performance-based design requires quantitative analysis of various aspects and
factors that may affect total building performance. In fire safety engineering (FSE),
performance-based codes and practice have existed for more than a decade in New Zealand,
Australia, Canada, the UK, and Japan, etc (Bukowski and Babrauskas 1990; Snell et al.
1993; Richardson 1994).
However there is still a long way to achieve wide application of performance-based design
in FSE (Babrauskas 1996; Brannigan 1999). For example, Brannigan indicates that currently
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
12
scenario fires (design fires) are usually adopted in fire safety assessment rather than the use
of fire scenarios that closely resemble real building environments. This is a serious
limitation for assessing building performance in real world fires. Therefore, as identified in
Chapter 1, acquiring a solid knowledge of fire behaviour is still a major task for improving
the performance-based approach. This knowledge of fire behaviour covers fire origin
(thermal decomposition and ignition), development (combustion and flame spread) and
outcome (heat release and gaseous species yields).
Fire models, including mathematical, empirical and semi-empirical models, are
mathematical statements to represent our theoretical understanding about fire. They are used
to simulate or predict a process or phenomena in a fire environment, including fire
development and consequences (Cox 1994). Today almost all the fire models available are
computer programs. These computer programs perform large numbers of time consuming
and lengthy calculations and provide the user with a set of parameters that describe the
phenomena or events being simulated. Computer modelling of fire development, smoke and
fire spread is a desirable major component in risk assessment models. Fire growth models
are used to predict the fire growth characteristics of various fire scenarios in an apartment.
The predictions of particular aspects of hazard development include the burning rate, room
temperatures, oxygen and toxic gas concentrations, etc. Such data are critical for a risk-cost
system model to accurately estimate the expected risk-to-life and fire-cost expectations.
The development of fire models shows progress from simple to complex. The complexity of
models has two meanings: detailed phenomena with a smaller magnitude and more aspects
of problems being considered and evaluated.
As for the physical magnitude of combustion phenomena, the fire models have progressed
from zone models to field models. The zone models divide the whole enclosure into two
zones (an upper hot gas layer and a lower cool fresh air layer) and a plume may be also
modelled to bring combustion products and entrained air from the lower zone to the upper
zone. The thermal properties within these two zones are treated as uniform. The size of the
zones may exceed a couple of metres. The field models, which are also called as
computational fluid dynamics (CFD) models, divide the enclosure into thousands of cells.
The magnitude of the cells can be as small as millimetres, depending on the computational
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
13
capability and duration applied to the models. Brief introductions for these two types of fire
models will be given in Section 2.2.
As mentioned in Chapter 1, a characteristic mesh size required for a CFD model to capture
the details of chemical reaction zone in a fire may be smaller than a magnitude of
millimetres. Therefore to study the pyrolysis and ignition processes in details, including gas
phase mixing and reaction, CFD models are appropriate and available. A CFD model was
then chosen for this research purpose.
As for the fire phenomena embedded within the combustion models, the sophistication of
the models has grown rapidly in the past decades. Initially, these models were only able to
describe limited phenomena of fires that were observed in fire growth. Gradually, more and
more aspects of fire were added. A full picture with certain details for the fires is now
appearing (Xue et al. 2001). For example the treatment of the fire source in these models
developed as following, according to Friedman’ s international survey of fire models
(Friedman 1992):
Set, in the simplest case, a heat release source and specify the fire base area and
pyrolysis rate of the combustible
Specify a fire with heat release rate varying in a prescribed manner with time
Add the influence of oxygen concentration on combustion into the calculation
according to some formula
Take radiative feedback into the burning rate calculation
Set the burning rate according to full-scale experimental results, while bench-
scale data can be adopted to predict relevant burning characteristics of the
combustibles
Today the treatment of the fire source in fire models has become even more mature. In some
complicated combustion models, fire development and outcome are estimated based on
computations of complex heat transfer and flame spread rather than simply being specified
by users (Atreya et al. 1986; Quintiere 1993; Staggs and Whiteley 1999). It is found that a
great number of basic quantities are needed to describe more and more phenomena in
current fire models. Delichatsios and Saito (1991) provide a list of the quantities needed for
a scientific fire model of upward flame spread over a charring surface:
The ignition parameters (intensity, size, duration)
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
14
Thermal conductivity, density and heat capacity of the virgin material and of the
char
Surface temperature of the pyrolyzing surface
Surface reflectivity
Heat of gasification
Heat of combustion of pyrolysis gases
Combustion efficiency
Radiative fraction of flame heat output
It should be noted that this list is not completely definitive. Some other factors/quantities,
such as heat of combustion of char and the effect of composite material, may need to be
taken into consideration as well.
To achieve an engineering approach for modelling combustion, it is necessary to identify
the input quantities and phenomena that are most relevant to fire initiation, development and
outcome. Then these quantities and phenomena must be included in models to obtain
desirable engineering results. Note that over time, more and more details will be added into
these models with an acceptable modelling duration and accuracy. These efforts will enable
fire models to deal with more complex and realistic fire situations. Detailed reviews on
research activities about pyrolysis, ignition, flame spread over solid fuel are performed in
Sections 2.3 to 2.5. Research requirements in those areas for the current study are then
identified.
Based on the results of the theoretical study, a series of test methods and devices were
developed to obtain the basic properties of materials and to investigate the burning
behaviour and fire performance of building and their components. These experimental data
are also critical for validation of modelling results. The experimental equipments range from
small-scale/bench-scale (such as the TGA, the cone calorimeter and Lateral Ignition and
Flame Transport tests) to middle-scale (furniture calorimeter and ISO room tests) and full-
scale sizes (doorway and room calorimeter tests). The basic quantities obtained from the
small/bench-scale tests are normally used to predict material or component performance in
the full-scale tests through theoretic analysis and modelling. The method for predicting
ignition time developed by Quintiere and Harkleroad (1984) is such an example.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
15
Since the research interest in the current study is to describe the pyrolysis and ignition
processes in detail, only experimental methods relevant to this purpose, normally the
small/bench-scale tests, are reviewed. Their major applications are discussed in Section 2.6.
2.2 Fire Models
During the past two decades, a number of reviews on fire models have been made by
various researchers. Some of the reviews focused on a comparison and evaluation of the
overall ability of the models, like those by Jones (1983), Friedman (1992), Dembsey et al.
(1995) and Beard (1997). Other reviews concentrated on details of modelling schemes for
special phenomena, such as: effect of combustion (Xue et al. 2001), pyrolysis of char
forming solid (Moghtaderi 2001), the role of condensed phase in polymer combustion
(Moghtaderi 2001), and ignition of wood (Babrauskas 2001), etc.
In this literature review, it is not intended to perform an exhaustive review of all the fire
models available. However a general description of major types of fire models is given in
this chapter. In the next chapter, emphasis is laid on those models related to the current
study: pyrolysis and ignition of solid fuels, especially of some types of furnishing materials.
The principle kinds of deterministic models that have been developed are zone and field
models. A brief review of these two kinds of models is given in the following sub-sections.
2.2.1 Zone models
A modelling approach for predicting various aspects of fire phenomena in enclosures has
been called “ zone” modelling. It is based on a conceptual representation of the compartment
fire process, and is an approximation to reality. The zone model simply represents the
system as two distinct compartment gas zones (Peacock et al. 1993):
an upper, hot volume and
a lower, cold volume
resulting from thermal stratification due to buoyancy.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
16
Conservation equations (for energy, mass and momentum transport) are applied to each
zone and serve to embrace the various transport and combustion processes that apply. The
fire is represented as a source of energy and mass, and manifests itself as a plume, which
acts as a “ pump” for the mass from the lower zone to the upper zone through a process
called “ entrainment.”
The properties of the upper and lower zones are assumed to be spatially uniform, but can
vary with time. Thus, temperature, T, and species mass concentration, Yi, are properties
associated with ideal upper and lower homogeneous layers. The gases in the layers are
treated as ideal gases with constants of specific heats cp and cv. The pressure in the
enclosure is considered uniform in the energy equation, but hydrostatic variations account
for pressure differences at free boundaries of the enclosure.
The solution process of the conservation equations is completed by each source or transport
term being given in terms of the layer properties. The extent to which source and transport
relationships are included reflects the sophistication and scope of the zone model. These
source and transport relationships are usually termed “ sub-models” and can be subroutines
in the computer code of zone model. Due to the limited knowledge of fire behaviour, many
assumptions are adopted and many of the sub-models still need further improvement. For
example, the mass of fuel supply can be a result of a fire spreading over an array of different
solid fuels. However, the fuel properties are still not completely defined or conventionally
accepted for fire applications. The main reason for this is that no general exhaustive theory
exists for pyrolysis, and theories of flame spread and ignition are couched in terms of
effective fire properties. Meanwhile the treatments for mass and heat transports (such as
vent flows, various heat transfer, mixing of layers) are still very simple and sketchy in
today’ s zone models due to complex geometry and fuel configuration (Tanaka 1983;
Peacock et al. 1991).
A detailed review of the zone models, including the conservation equations and the sub-
models can be found in (Quintiere 2002).
The beginnings of zone fire modelling were marked by the fundamental equations published
by Quintiere in the mid-1970s (Quintiere 1977). The first pre-flashover zone fire model
published was PFIRES by Pape et al. (1981). The next publicly accessible model was
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
17
HARVARD by Emmons and Mitler (1985). At about the same time the first zone fire model
written specially for IBM-compatible personal computers was introduced by Walton (1985).
Several zone models have been developed at the Building Fire Research Laboratory, USA.
A summary of those typical zone models is given by Jones (2000). These models include
FAST, FIRST, ASET, CCFM, CFAST, FPETool and HARZARD. CFAST is a good
example in the way it represents the underlying physics of zone models.
CFAST is capable of predicting the environment in a multi-compartment structure subjected
to a fire. It solves conservation equations of mass and energy by using the ideal gas law and
relations for density and internal energy. The predictions obtained from these equations are
functions of time quantities, such as pressure, layer heights and temperatures. The model
calculates the time evolving distribution of smoke and fire gases and the temperature
throughout a multi-compartment building during a user-specified fire (Jones and Forney
1990).
An important limitation of CFAST is the absence of a fire growth model. The model utilizes
a user specified fire, expressed in terms of time specified rates of energy and mass released
by the burning item(s). Such data can be obtained by measurements taken in large- and
small-scale calorimeters, or from room burns. However, there are limitations in associating
such data for modelling real compartment fires. For example the data obtained from a
furniture calorimeter is derived from a “ free burning” test, and the test environment is
different from real compartment fire conditions (such as radiation feedback from hot upper
layer gases) (Jones 2001).
Some major features of CFAST are briefly described as follows (Jones 2001):
Fires: treated as fuel source with a specified releasing rate of combustibles, and
the burning rate controlled by oxygen concentration
Plumes and Layers: a plume model is adopted to predict plume entrainment
including mass and enthalpy transferring
Vent Flow: two types of vent flows, horizontal and vertical, are determined by
pressure difference across a vent based on Bernolli’ s law
Heat Transfer: material thermophysical properties are assumed to be constant for
convective and radiative heat transfer. This can cause error in gas phase radiative
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
18
calculations since the emissivity is a function of the concentration of species of
some strong radiators that change as the fire develops
Species Concentration and Deposition: a combustion chemistry scheme based
on a carbon-hydrogen-oxygen balance is used to estimate the rate of production
of species
It can be seen from the above reviews that zone models are not suitable for the current
research purpose. The major limitations of zone models are the lack of descriptions of the
fire source and flame spread process, and low accuracy in the computation of mass and heat
transfer, especially under complex geometry and environment conditions.
2.2.2 Field models
Field models are complex fluid mechanical models of turbulent flow derived from classical
fluid dynamics theory. This type of model solves versions of the fundamental equations of
mass, momentum, and energy (Stroup 2002). In order to facilitate the solution of the
equations, the compartment is divided into a three-dimensional grid of cells. The field
model calculates the physical conditions in each cell as a function of time. The calculations
account for physical changes generated within each cell and changes in the cell resulting
from changes in surrounding cells.
2.2.2.1 Computational Fluid Dynamics (CFD) basics
Classical fluid dynamics is concerned with the mathematical description of the physical
behaviour of fluids (gases or liquids). The equations consist, in general, of a set of three-
dimensional, time-dependent, non-linear partial differential conservation equations, referred
to as the Navier-Stokes (N-S) equations.
The N-S equation describing the conservation of mass is described by the continuity
equation.
( ) SUt
=•∇+∂∂ ρρ
(2.1)
where:
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
19
ρ: local density of fluid, kg/m3
U: fluid velocity vector with components in x, y, and z directions, m/s
S: sources or sinks of mass, kg/m3⋅s
t: time, s.
Conservations of momentum and energy are given as below, the equations (2.2) and (2.4)
respectively:
( ) σρρ ∇+=•∇+∂
∂BUU
tU
(2.2)
where:
( )[ ]TUUp ∇+∇+−= µδσ (2.3)
and,
B: body force, here is due to gravity (buoyancy), N
p: pressure, N/m2
: boundary thickness, m
: molecular viscosity, kg/m⋅s
T: temperature, K
( ) ( )tp
TUHtH
∂∂=∇∇−∇+
∂∂ λρρ
(2.4)
where:
H: total enthalpy, J/kg, given as a function of static enthalpy, h, by
H = h + ½U2
λ: thermal conductivity, W/m⋅K
Analytical solutions of these equations exist for a limited number of special cases that have
simple boundary conditions. In their most general form, the N-S equations can not be solved
by analytical methods. Therefore, solving the equations usually requires the use of
numerical techniques. Computational fluid dynamics (CFD) involves the numerical solution
of the N-S equations using computers. Field models solve the fundamental partial
differential equations of motion and conservation numerically at a discrete moment in time
and point in space. Using a set of grids in three dimensions, the compartment under study is
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
20
divided into many small volume elements or cells. The governing differential equations are
solved simultaneously for each cell to obtain various parameters.
2.2.2.2 Turbulence
The flows occurring in room fires are turbulent, generating eddies or vortices of many sizes.
The energy contained in large vortices cascades down to smaller and smaller vortices, until
it diffuses into heat. Eddies exist down to the size where the viscous forces dominate over
inertial forces and energy is dissipated into heat. For typical fires, this scale is of the order of
a millimetre or so. Using cells of this tiny scale would result in problems that cannot be
handled by most computers. As a result, turbulence models have been developed to account
for the effect of small-scale fluid motion on the motion in large-scale cells. A turbulence
model estimates the effect of small-scale or sub-grid phenomena on motion in the larger
scale.
Several turbulence models have been developed. The traditional approaches solve the so-
called Reynolds-averaged Navier-Stokes (RANS) equations (Cox and Kumar 2002). The
common type of model adopting the format of the RANS equations is the eddy viscosity
model which specifies the Reynolds stresses and fluxes algebraically in terms of known
mean quantities. The k-ε models, probably the best known and most widely used turbulence
models, are examples of eddy viscosity models (Launder and Spalding 1974). This type of
model has two additional differential equations per control volume. The first equation
governs the distribution of the turbulent kinetic energy k, while the second describes the
dissipation of the local energy ε. The equations used in the k-ε model contain several
empirical constants, which should be verified when used for new applications to ensure that
reasonable results are obtained. Currently, most CFD models use k-ε models to handle the
turbulence flow.
An alternative approach is to exploit Large Eddy Simulation (LES) techniques, where the
larger turbulent vortices are simulated rigorously (McGrattan et al. 1998; Zhou et al. 2000).
However this approach does not capture the length and times scales associated with the
details involved in combustion. Another methodology, direct numerical simulation (DNS),
that solves the exact equations rigorously is still under development (McGrattan et al. 1998;
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
21
Vervisch and Poinsot 1998). It needs much larger computer resources and more detailed
combustion-reaction information. Generally, with the current widely available computer
hardware, simulation involving direct integration of the full Navier-Stokes equations in
three dimensions is still not practical.
A detailed review of CFD models can be found in an article written by Cox and Kumar
(2002). In that paper, the authors discussed many of the issues involved with a CFD model,
such as turbulent flow, combustion, radiant heat transfer, boundary conditions, numerical
solution method, validation and some practical applications.
In the next chapter of this thesis, some typical CFD models developed to model pyrolysis
and combustion processes will be discussed in detail.
In summary, CFD models are able to simulate complex phenomena in combustion with a
relatively small cell size. Adopting some form of appropriate numerical solutions, such as
the k-ε models or the large eddy simulation technique for the turbulence problem, accurate
results can be obtained within an acceptable computational duration.
2.2.3 Comparison between the zone and field models
Zone models are relatively simple, running rapidly and cheaply. Zone models have the
ability to reduce computational complexity of fire growth and smoke spread modelling with
a certain level of accuracy. This makes a zone model a handy tool for investigating the
hazard of fire in buildings, which deals with hundreds of variables.
Zone models utilize many equations employing empirical relationships and constants
obtained from experiments. There are three serious limitations of zone models. The first is
that the empirical expressions used to describe physical behaviour in zone models may
break down under certain circumstances (e.g., as the geometries become more complex.)
The second limitation is that the scale of the combustion behaviours is much smaller than
the zones in the zone models, and then the combustion is not properly modelled. The third
limitation is that zone models treat each zone as being uniform, and no details within each
zone can be analysed.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
22
On the other hand, field models have the potential to be applied to a variety of problems
with only minor modifications. Field models are able to estimate detailed velocity and
temperature distributions, whereas the zone models only calculate average or maximum
temperatures and velocities for a few points in a room. When study interests focus on
detailed phenomena, field models are the perfect choice for the modelling tools. For
example, plume entrainment, pyrolysis process and flame spread on a solid are just few of
the areas that create huge difficulties for the zone models.
Fluid dynamic considerations are automatically built into field models to avoid
oversimplified approximations. Field models are used in analysing problems involving far-
field smoke flow, complex geometries (e.g., sprinkler links), and impact of fixed ventilation
flows (Stroup 2002). Field models have been successfully used to predict special fire
behaviours, such as the “ trench effect” (Cox et al. 1989) and low temperature “ islands”
(Beard 1992).
This does not mean, however, that there are no unresolved difficulties with this kind of
modelling. The application of field models to real fires is still at the preliminary stage,
though significant progress has been achieved recently (Babrauskas 2002). The modelling
of the complicated combustion process of solids, including the surface temperature and
pyrolyzing flux, the near field entrainment and flow, and the process of transition from solid
to gas phase, is yet to be adequately incorporated into field models. Meanwhile the
simulation accuracy is greatly dependent on a better understanding of various phenomena in
detail and appropriate selection of properties for simulated materials.
2.3 Thermal Decomposition
It is necessary to distinguish three commonly confused concepts: thermal decomposition,
thermal degradation and pyrolysis. According to definitions given by the ASTM
(ASTM_E176), thermal decomposition is:
“ a process whereby the extensive chemical species change caused by heat,”
and thermal degradation is:
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
23
“ a process whereby the action of heat or elevated temperature on a material,
product, or assembly causes a loss of physical, mechanical, or electrical
properties.”
Pyrolysis is a process of chemical decomposition of solid and gaseous products released by
effort of heating. In other words, pyrolysis is a special case of decomposition in which
gaseous products are released from the decomposition of solid materials.
In a fire situation, thermal decomposition is an important change, which occurs before
ignition of solid materials. This decomposition involves physical and chemical changes that
usually generate gaseous combustibles vapours (pyrolysis). Normally physical changes,
such as melting and charring, can markedly alter the composition and burning
characteristics of a material, while chemical processes are responsible for the generation of
flammable volatiles. Beyler and Hirschler (2002) provide a detailed description of the
physical and chemical processes involved in the decomposition of a polymer.
The thermal decomposition of some identified furnishing material will be investigated in the
current research. Major processes affecting the decomposition will be identified. Methods
used to describe the decomposition by thermal dynamics will be discussed. All such
knowledge is useful for determining the ignition of solid fuels and fire development, since
the thermal decomposition affects not only the conditions for ignition but also the supply of
combustibles to the fire.
2.3.1 Physical and chemical processes
Solid fuel materials usually undergo both physical degradation and chemical decomposition
changes when heat is applied to them. When thermal decomposition takes place a solid
material generates gaseous fuel, which can burn above the surface of the material. In order
for the process to be self-sustaining, it is necessary for sufficient heat to be fed back to the
material to continue the production of gaseous fuel vapors or volatiles. This heat transfer
normally comes from the burning gases as well as possible external heat sources.
The various physical processes that occur during thermal decomposition depend on the
nature of the material. The common processes are: softening of thermoplastics, melting of
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
24
crystalline materials, deformation, etc. Cellulosic materials have been well studied due to
their wide application. These materials, due to their structure, may not change state below
the temperature at which thermal decomposition occurs. Since water is absorbed physically
and chemically, release of water is a change that always occurs during a heating process
with a substantial temperature rise. The activation energy for physical desorption of water is
30 to 40 kJ/mol, and the physical desorption starts at temperatures somewhat lower than
100 oC (Beyler and Hirschler 2002).
Another noticeable phenomenon associated with thermal decomposition of some materials
is charring. Chars can be formed on many materials, such as cellulose, thermosetting
materials and thermoplastics. The physical structure of these chars will strongly affect the
thermal decomposition process (Goos 1952). Low density and high porosity chars are good
thermal insulators and can significantly inhibit the flow of heat from the combustion flame
back to the condensed phase behind it. As the char layer becomes thicker, the heat flux to
the un-charred part of the materials decreases, and the thermal decomposition rate is
reduced.
The thermal decomposition of solid materials may proceed by oxidative processes or simply
by the action of heat. In many solid materials, the thermal decomposition processes are
accelerated by oxidants, such as air or oxygen. The minimum decomposition temperatures
in the presence of an oxidant are lower than that without the oxidant. Therefore the
concentration or presence of oxygen is very important in determining the thermal
decomposition rate and mechanisms. The effect of oxygen makes the prediction of the
thermal decomposition rate much more complicated, since the prediction of the
concentration of oxygen at the solid surface during thermal decomposition or combustion is
quite difficult. Some efforts have been made to study the effect of oxygen concentration on
the decomposition process (Kashiwagi et al. 1985; Brauman 1988; Gijsman et al. 1993).
Kashiwagi and co-workers (1985; 1992) found that a number of material properties affect
the thermal and oxidative decomposition of thermoplastics. These properties include
molecular weight, prior thermal damage, weak linkages, and primary radicals. Their
research also resulted in the development of models for predicting the kinetics of general
random-chain scission reaction, as well as for the thermal decomposition of celluloses and
thermoplastics.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
25
There are numbers of general chemical mechanisms to describe thermal decomposition of
polymers. These can be listed as follows (Beyler and Hirschler 2002):
random-chain scission
end-chain scission
chain-stripping
cross-linking, etc
Thermal decomposition of a solid material generally involves more than one of these classes
of reactions. These general classes may only provide a conceptual framework useful for
understanding and classifying solid material decomposition behaviour.
Some materials, especially thermosets and cellulose, have more complex decomposition
mechanisms. Indicated by previous researchers (Beyler and Hirschler 2002), polyurethane
(particularly flexible foams) can be decomposed by three different mechanisms. One of
them involves the formation of gaseous isocyanates, which can then re-polymerize in the
gas phase and condense as a yellow smoke. Cellulose, such as wood, decomposes into three
types of products:
laevoglucosan, which quickly breaks down to yield small volatile compounds
a new solid, char
a series of high molecular weight semi-liquid materials, known as tars
Since the decomposition processes of polymers are so complicated and there are a great
number of unknowns in the processes, identifying the major processes and products is a
challenge for the study of decomposition. It is especially true for the kinetic description of
the thermal decomposition.
2.3.2 Kinetics
There are several different usages of “ kinetics” . Generally, “ kinetics” means chemical
kinetics, i.e. the reaction rates of chemical processes. In order for a reaction to occur, a
certain threshold that should be reached is called the “ activation energy” . This activation
energy, and other parameters to describe a reaction, such as a pre-exponential factor, are
commonly called as “ kinetics” by many researchers. This is also the usage in this thesis.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
26
Normally a chemical reaction rate between two reactants, A and B, can be expressed as
follow (Beyler and Hirschler 2002):
[ ] [ ]nm BAkratereaction = (2.5)
where:
k: the reaction rate constant, 1/s;
[A] and [B]: are concentrations of the reactants A and B, mol/m3;
m and n: orders of reactions with respect to A and B respectively.
There may be up to eight generic types of reaction involved in decomposition processes.
The kinetics describing the process can be quite complex. It is possible to give an
expression for the rate of weight loss of solid fuel caused by some relatively simple
chemical reactions, such as random-chain initiation and end-chain initiation (Beyler and
Hirschler 2002). A simple expression of the rate of weight loss for the end-chain initiation is
given as:
knD
dtdW
p ⋅⋅= )2( (2.6)
where:
W: weight;
Dp: degree of polymerization, or number of monomer units per polymer chain;
n: is the number of polymer chains;
k: rate constant for end-chain initiation.
However, the reaction rate constant and reaction order for the expressions are somehow
difficult to decide.
From an engineering point of view, the above chemical expressions for the reaction kinetics
are either unsolvable or inapplicable. Therefore, for engineering applications, some
simplified kinetic descriptions are considered. The general simplification is to use an overall
(global) decomposition process to represent the possible multiple reactions existing in the
decomposition process of the polymer. This overall decomposition can be one-step
(representing one major reaction) or multiple steps (representing a multiple reaction).
Moghtaderi (2001) carried out a detailed review of these overall decomposition models.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
27
Details for application of the overall decomposition models will be given in the next chapter
where modelling for pyrolysis and ignition processes is described.
The simplest and most commonly adopted assumption for utilizing the simple overall
kinetics expressions is that the reactions can be described by a one-step first order Arrhenius
expression:
RTE
Aek−
= (2.7)
where:
k: the reaction rate constant, 1/s;
A: the pre-exponential factor, or frequency factor, 1/s;
E: activation energy, kJ/mol;
R: universal gas constant, = 8.314×10-3 kJ/mol⋅k;
T: reaction temperature, K.
For example, Kashiwagi and Nambu (1992) obtained the overall kinetic constants for the
thermal oxidative decomposition process of a cellulosic paper. The predicted decomposition
rate matches with the experimental result reasonably well.
The simplified overall kinetic approach, which uses the overall kinetics to describe the
reaction, has been adopted for many decomposition models. The key point is how to define
the overall reaction and how to obtain the kinetics. A standard kinetic test method is the
thermogravimetric analysis (TGA) test, which will be briefly introduced late in this chapter.
The efforts to improve the experimental and calculation scheme for the kinetics are further
discussed in Chapter 4 where TGA tests are carried out for the studied materials.
2.3.3 Decomposition of the furnishing materials
As mentioned in Chapter 1, furnishing materials play an important role in ignition and
combustion of fires in compartments. Six furnishing materials have been chosen for the
current research to represent typical materials used in furniture manufacture and inner
furnishing. Two are timber: mountain ash (botanical name Eucalyptus regnans) and
Australian pine (botanical name Casuarina equisetifolia); two are fabrics: cotton fabric and
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
28
cotton and polyester fabric; and two are polyurethane (PU) foam, standard PU foam and
Stamina PU foam. There are a number of papers which report the thermal properties and
combustion behaviour of these materials. For cotton materials, these are (Cleary et al. 1992;
Ohlemiller and Shields 1995), for polyurethane foam (Luo and He 1997; He et al. 1999;
Walther et al. 2000), and for wood (Atreya et al. 1986; Li and Drysdale 1992; Moghtaderi et
al. 1997). However, the pyrolysis of wood is a very complex process due to its composition.
It is necessary to discuss various stages of decomposition process of this material
individually.
2.3.3.1 Pyrolysis and charring of wood
The physical and chemical changes that occur in the pyrolysis of wood can mainly be
considered as charring and gasification. However the reactions involved may be quite
complicated. As described by Beyler and Hirschler (2002), there are at least four processes
in the decomposition of celluloses, the major component of wood:
1. cross-linking of cellulose chains with the evolution of water (dehydration)
2. unzipping of the cellulose chains to form laevoglucosan from the monomer unit
3. decomposition of the dehydrated products (dehydrocellulose) to yield char and
volatile products
4. further decomposition of the laevoglucosan into smaller volatile products,
including tars, and finally carbon monoxide
Meanwhile a simple desorption of bound water may be also included in the decomposition
process of wood.
Wood consists of approximately 50% cellulose, 25% hemicellulose, and 25% lignin. The
decomposition of wood may be treated as the combined decomposition mechanisms for the
individual components. It can be observed from the yields of volatile products and kinetic
data. When heated, the decomposition temperature ranges for the hemicellulose, cellulose
and lignin are 475 to 535 K, 525 to 625 K, and 555 to 775 K respectively. Therefore the
decomposition kinetics of a specific species of wood will be determined by its composition.
It is also known that the decomposition of lignin contributes significantly to the overall
yields of char, and that the lignin fraction supports most of the subsequent glowing
combustion. The cellulose fraction contributes most to flaming combustion (Tang 1972).
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
29
In the gasification of the solid, only part of the original fuel becomes a vapor and a certain
amount solid residue is left. If the residue is carbonaceous, the char is formed on the surface
first. When the thermal decomposition of deeper layers of the solid continues, the volatiles
produced must pass through the porous char above. The hot upper char layer can cause a
secondary reaction in the volatiles. The top char layer can be burnt by an oxidation reaction
at a higher temperature. On the other hand, the forming of top char layer can prevent the
underlying layers from further thermal breakdown or at least slow the process down. The
combination of these factors will be affected by heating conditions and thermal properties of
the wood.
Many factors are involved in wood charring. Kanury and Blackshear (1970) checked
various physiochemical effects, including diffusion of condensable vapors, internal
convection outward, properties of the partially charred wood, kinetics of pyrolysis, and post-
decomposition reactions. Lee et al. (1977) added other factors to the wood charring, such as
external heating rate, total time of heating, and anisotropic properties of wood and char
which related to internal flow of heat and gas. Among the factors some characteristics of
wood greatly affect the charring rate. These characteristics are density, moisture content,
permeability, and chemical composition. Density is recognized as the major factor affecting
charring rate. According to Schaffer’ s study (Schaffer 1967), denser wood has a slower
charring rate. A similar relationship was also observed for wider wood species by other
researchers (Hall et al. 1972). Permeability may be a controlling factor in the movement of
moisture, which also affects the outflow of pyrolysis gases (Kanury and Blackshear 1970).
The pressure in a heated wood was measured and reported to drop when structure changes
were observed (Roberts 1970). The relationship between inner pressure and external
radiative flux has been measured and included in Fredlund’ s theoretical model (Fredlund
1988).
The gaseous pyrolysis products of wood are a complex and highly variable mixture. For
instance, the volatile materials detected as products of wood decomposition have been
reported as being from 30 up to 200 compounds (Goos 1952; Beck and Arnold 1977). By
investigating the composition of pyrolysis products of wood, Abu-Zaid (1988) and
Nurbakhsh (1989) found that the total hydrocarbons and carbon monoxide are about 25%
(mass fraction) of the total pyrolysed gas. For modelling purposes, the volatiles are normally
represented by four major gases H2, CO, CH4, and CO2 (Klose et al. 2000). Of course this is
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
30
a certain degree of simplification. More detailed knowledge about the pyrolysis products
may be still required.
2.3.3.2 Role of moisture content
Changes of moisture content in a solid piece of wood involve vaporization and condensation
processes. These changes may affect the pyrolysis process greatly, as explained below.
The moisture content in wood exists in two basic forms: (1) bound or hygroscopic water,
and (2) free or capillary water. For most wood species the moisture content at the Fibre
Saturation Point (FSP) is about 30% while the total moisture content could be as high as
60%. When exposed to fire, the temperature of the wood will rise to a point, near 100 oC,
when the moisture starts to evaporate. Since the bound water is adsorbed into the cell walls
(at least if the moisture content is below the FSP), evaporation for this portion of water
requires more energy than needed to boil free water. During this initial period, most of the
heat received by the fuel is consumed by heating and evaporating the free water portion.
When the total moisture content of the surface region drops to a level close to the FSP, the
evaporation front begins moving into the solid. The water vapor largely migrates toward,
and escapes through, the exposed surface. A fraction also migrates in the opposite direction,
and re-condenses at a location where the temperature is below 100 oC. The movement of the
moisture within the wood has a significant impact on the heat and mass transfer processes
taking place within the solid. For example the heat conductivity may increase several times
compared to when the wood has a completely “ dry” status.
The dryer wood portion further increases in temperature until the fibres begin to degrade.
The thermal decomposition of wood starts around 200 to 250 oC. The volatiles that are
generated again travel primarily toward the exposed side, but also partly in the opposite
direction. They consist of a mixture of combustible gases, vapors of moisture and tars. A
solid carbon char matrix remains. The volume of the char is smaller than the original
volume of the wood. This results in the formation of cracks, bends and fissures which
greatly affect the heat and mass transfer between the flame and the solid. The combustible
volatiles that emerge from the exposed surface mix with ambient air and burn in a luminous
flame.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
31
Generally, the moisture content and its movement may significantly affect the pyrolysis
progress in two aspects: (1) changes in the thermal properties; and (2) overall heat transfer
and balance, such as heat of vaporization.
2.4 Ignition Controlling Mechanism
The flaming ignition of a combustible material, the subsequent spread of flames over its
surface and the establishment of steady burning are critical processes in the development of
a fire. They determine the rate of growth of the fire and the rate of heat release.
A fundamental understanding of these processes is basic to the development of material
flammability and combustion test methods. Such tests are used to provide meaningful and
reliable information about the behaviour of the material in an actual fire.
There are several ways that a solid combustible can be ignited. The ignition, or combustion
reaction initiation of a combustible material, can occur either in the solid or in the gas
phases. In fires, the latter (flaming ignition) is the most important since it may lead to the
spread of the fire.
The gas phase ignition of a solid combustible is generally a combination of a number of
events. The first is exposure to an externally imposed heat flux (radiation and/or convection)
that causes gasification of the solid. The second is the presence of conditions (in the gas or
external to it) that will lead to the onset of a sustained combustion reaction between the
vaporized fuel and the oxidizer gas. If the reaction is initiated by an ignition source (such as
open flame, electrical spark, etc) the ignition is normally referred to as piloted ignition. If
ignition occurs without a pilot the process is normally referred to as spontaneous or auto-
ignition. For this latter type of ignition to occur, the gaseous oxidizer/fuel mixture must be
at an elevated temperature. Gas phase ignition caused by a high surface temperature or by a
surface reaction, such as smouldering or char oxidation, can also be included in this
category.
A detailed description for these two types of controlling mechanisms was given by
Fernandez-Pello (1995). Brief descriptions of them are in the following sections.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
32
2.4.1 Auto Ignition
The auto ignition study was based on a simplified experimental ignition model provided by
Niioka (1981). See Figure 2.1.
Figure 2.1 One-dimensional experimental ignition model
The physical description of the simplified ignition model can be given as follows. A solid
fuel slab with a thickness s and a constant initial temperature TI is suddenly exposed to a hot,
stagnation point gas flow. This hot gas flow consists of oxidizing gas with properties
marked as ambient parameters, U∞, T∞, and Yoxy, ∞. This works as an external heat flux and
the heat flux is absorbed at the solid surface, which heats the solid and eventually causes it
to pyrolyse. The pyrolysed fuel vapour convects and diffuses outwards, mixes with hot
oxidizer gas and forms a combustible mixture near the surface. The high gas temperature
helps initiation of a gas combustion reaction. If the energy absorbed from the hot gas is high
enough to overcome the heat loss to the surroundings, a sustained combustion reaction
(flame) will occur over the combustible surface.
By studying various flow velocities and distances to the solid surface, Niioka suggested that
two primary mechanisms control the solid fuel ignition process. One is the heating and
gasification of the solid and the other is the onset of the gas phase chemical reaction. Some
times (t) related to this ignition process are defined as follows:
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
33
the ignition delay time, also shorten as ignition time, tig, is the time from the
sudden exposure of a surface to a hot gas flow at a stagnation point to the onset
of combustion
solid pyrolysis time, tpy, related to the heating and pyrolysis of the solid, is the
time from the beginning of exposure to the hot gas to the starting of pyrolysis on
the solid surface (at pyrolysis temperature Tpy, (Carslaw and Jaeger 1959))
gas induction time, tin, related to the onset of the gas phase chemical reaction and
also named as gas phase time delay, is the time from starting of pyrolysis of the
solid fuel to the ignition.
The gas phase time delay is then decided by two times:
The flow residence time, which is the time for pyrolysed gas to reach the
boundary layer edge, and
the chemical time, which is the reaction time of the pyrolysis products in the gas
phase.
The solid (total) ignition delay time is the sum of the solid pyrolysis time and the gas
induction time:
inpyig ttt += (2.8)
For example, as the flow velocity is increased the surface heat flux also increases and
consequently the time required to heat up and pyrolyse the solid decreases. The influence of
this mechanism is graphically represented by the descending pyrolysis time line in Figure
2.2. On the other hand, it is noticed that an increase in the gas velocity results in a decrease
of the flow residence time, if compared to the chemical time. This will delay the onset of the
chemical reaction and eventually prevent its initiation. The effect of this mechanism can be
graphically represented by the ascending line, i.e. the induction time. Therefore, the solid
ignition delay time can be decided by the sum of the solid pyrolysis time line and the gas
induction time line.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
34
Figure 2.2 Solid ignition controlling mechanisms and ignition time variation
If the gas phase chemical time is very short compared with the flow time, the induction time
will also be very small and the solid ignition delay will be controlled by the heat transfer to
the solid. This will occur for low flow velocities, and/or high temperature and oxygen
concentrations. For high velocities and/or low temperatures and oxygen concentrations, the
solid ignition delay will be dominated by the onset of the gas phase chemical reaction.
These concepts can be extended to other geometrical configurations and other sources of
ignition. For example, the presence of a pilot flame affects the ignition process by locally
reducing the gas induction time. The solid heating is consequently the dominant controlling
mechanism of the solid ignition process, unless the flow velocity is very large or the oxygen
concentration is very low.
2.4.1.1 Solid heating and gasification
Since the oxidizer supplying flow in Figure 2.1 is uniform on the x-axis and the studied part
of the solid is limited in the middle of the flow, this physical problem can be treated as one-
dimensional. Fernandez-Pello (1995) gave a detailed analysis for this problem and provided
the following assumptions and equations. Assumptions adopted in the problem are:
solid properties are constant
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
35
the pyrolysis occurs at the surface of the solid
solid pyrolysis is described by a first order Arrhenius law
the radiation is absorbed and emitted at the surface
there is no surface oxidation or material charring
The governing solid phase equation for energy is:
vss
sss
sss LmT
yT
dtds
ct
Tc ''2 −∇=
∂∂+
∂∂ λρρ (2.9)
where:
mass flux is given by:
−
= sRTE
s eAm ρ'' , and
s: density of surface solid, kg/m3;
cs: specific heat of surface solid, J/kg⋅K;
Ts: surface temperature, K;
λs: thermal conductivity at surface, W/m⋅K;
Lv: heat of vaporization, kJ/kg;
As: pre-exponential factor for the pyrolysis reaction occurring at surface, 1/s;
E: activation energy, kJ/mol;
y: coordinate normal to the solid surface;
s: thickness of the solid, m.
Since the activation energy E is large for most materials, the pyrolysis rate is small for
surface temperature lower than a certain value (Williams 1985). Thus it can be assumed that
pyrolysis of the solid takes place at the surface and at a constant surface temperature Tpy,
which is defined here as the “ pyrolysis” temperature. For a one-dimensional process and for
the period prior to pyrolysis initiation (Ts<Tpy), Equation (2.9) is reduced to the heat
conduction equation
2
2
yT
tT
c ss
sss ∂
∂=∂
∂ λρ (2.10)
A solid is considered thermally thick if its thickness is larger than the thickness of the heat
penetration layer (at a particular time). It is considered thermally thin if the temperature
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
36
variation across its thickness is negligible. A theoretical judgement can be achieved by the
following equation, (Kanury 2002):
1)(/ , <<−′′ Iss TTLq λ (2.11)
where:
q ′′ : imposed flux, W/m2;
L: slab thickness, m;
λ: thermal conductivity, W/m⋅K;
Ts: surface temperature, K;
Ts,I: initial surface temperature, K.
If Equation (2.11) is satisfied, the solid may be considered thermally thin; i.e., a rapid
transfer of heat within a material would lead to establishment of a uniform temperature
profile. On the contrary, if the left hand of Equation (2.11), the ratio of imposed flux to the
conductive flux, is larger than 1, the solid body behaves as thermally thick. The latter can be
true either if the solid is physically thick or the thermal length, qTT Iss ′′− /)( ,λ , is small.
For thermally thick or thin solids, the pyrolysis time can be obtained from solution of
Equation (2.10) and the boundary conditions as follows:
for a thermally thick solid:
2''
2
4)(
s
Ipyssspy q
TTct
−=
ρπλ
(2.12)
for a thermally thin solid:
"
)(
s
Ipysspy q
TTLct
−=
ρ
(2.13)
where:
TI: initial solid temperature, K;
Tpy: pyrolysis temperature, K;
L: solid fuel thickness, m;
"sq : surface heat flux, kW/m2.
If the gas phase induction time is very small (low velocity and high oxygen concentration,
and high gas temperature or piloted ignition) the pyrolysis time will be nearly equal to the
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
37
solid ignition time. Equations (2.12) and (2.13) can be used to predict ignition times as a
function of the process parameters.
Because the fuel/oxidizer mixture becomes flammable almost immediately after solid
pyrolysis starts, the pyrolysis temperature is often defined as the “ ignition” temperature.
This definition, though not physically correct, can yield fairly accurate results.
For the thermally thick solid fuel, the solid heating process with the surface heat flux can be
expressed as follows, (Arpaci and Larsen 1984):
''''2/1'' )()/( radrspyPggs qqTTxuccq +−−= ∞∞ρλ
(2.14)
where:
"rsq : re-radiation from the solid surface, kW/m2;
)( 44"∞−= TTq pysrs σε
λg: gas thermal conductivity, kW/m⋅K;
g: gas density, kg/m3;
∞u : forced flow free stream velocity, m/s;
x: coordinate parallel to the solid surface (see Figure 2.1);
∞T : ambient temperature, K;
"radq : external radiation, kW/m2;
: Stefan-Boltzmann radiation constant, W/m2⋅K;
s: surface emissivity.
If it is assumed that the surface temperature is constant and equal to the pyrolysis
temperature and that the gas does not emit or absorb radiation, substituting Equation (2.14)
into (2.12) gives the solid pyrolysis time:
2""2/1
2
])()/([4
)(
radrspyPgg
Ipyssspy qqTTxucc
TTct
+−−−
=∞∞ρλ
ρπλ
(2.15)
If the heat fluxes at the surface of solid are balanced, or the surface is in thermal equilibrium
before the pyrolysis temperature is reached, ignition will not occur. The minimum heat flux
for ignition can be obtained by equating the denominator of Equation (2.15) to zero. This is
useful for determining a relationship between the minimum heat flux and ignition
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
38
temperature (assumed as the pyrolysis temperature), which was raised by Quintiere and
Harkleroad (1984) and discussed further in Section 2.4.2.
2.4.1.2 Gas phase reaction
When the ambient oxygen concentration is low or the gas velocity is high, the gas induction
time may be significant, and this has to be considered in the prediction of the solid ignition
times.
The theoretical modelling of the gas phase reaction includes the solution of the mass,
momentum, energy and species equations combined with the chemical rate equations.
However, an analytical solution for this problem is not easy to obtain. As discussed earlier
in Section 2.3, a common practical method to simplify the solution is the assumption that
the gas phase reaction is a single step overall reaction. Then the reaction rate can be
described by the Arrhenius equation:
−= RT
E
oxyFnFg eYYpAk
(2.16)
where:
k is reaction rate, 1/s, and
Ag: gas phase pre-exponential factor;
pF: pressure fraction of fuel;
YF: fuel mass fraction;
Yoxy: oxygen mass fraction;
n: reaction order number;
E: activation energy, kJ/mol⋅K;
R: universal gas constant, =8.314×10-3 kJ/mol⋅K;
T: gas temperature, K.
Williams (1985) developed expressions for the critical conditions of ignition and delay time
by using the property of large activation energies. Since a large number of the fuels
(particularly the hydrocarbon type) have large activation energies, also due to the character
of the Arrhenius combustion reaction, the reaction rate is very small for temperatures below
a critical value. Thus the solution of the problem can be divided into two solutions: one non-
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
39
reactive (frozen solution) and another active that applies to a very small region of the
mixture where the temperature is the highest.
A characteristic Damköhler number, Λo, is introduced to express the induction time.
ΛΓ−
−=
oin
dxdu
t 1ln4
1
(2.17)
where
−
∞
∞ ∞
∆=Λ RT
E
P
FsoxyIgRo e
dxdu
RTc
YEYnWHA
2
,
2
ρ
(2.18)
and
Γ = 4c(E/RT∞)[(2-β)/e2(1-β2)] (2.19)
β = T∞cP/YF,s∆Hc (2.20)
where all the variables have same meaning as in previous equations, and
cH∆ : heat of combustion, kJ/kg;
WI: initial solid weight, kg;
∞,oxyY : ambient oxygen mass fraction;
YF,s: surface fuel mass fraction.
The gas velocity gradient parallel to the solid surface, du/dx, is defined as stretch rate
(acceleration). It can be seen that as the stretch rate increases the induction time increases
and becomes infinite at the value of the stretch rate at which Λo =Γ. This also determines
the critical Damkohler number for ignition:
Λo,cr =Γ
If the Damkohler number is below the critical value, ignition will not occur. The common
reasons are a too low gas temperature, low oxygen concentration and large stretch rate. This
critical Damkohler number can be derived from steady-state analysis (Fernandez-Pello and
Law, 1982).
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
40
For large Damkohler numbers (Λo>>Γ) the gas phase induction time becomes inversely
proportional to the Damkohler number, tin=Γ/(4(du/dx)Λo) and becomes negligible. This
will occur mainly for high gas temperatures, moderate values of the flow velocity and
oxygen concentration. The presence of a pilot flame could be sufficient for this limit
because the reaction rate is very sensitive to temperature and only a little fuel vapor is
needed to initiate the reaction.
2.4.1.3 Ignition time, tig
For the simple situation discussed in Section 2.4.1, the solid’ s ignition delay time (ignition
time tig), is given by the sum of the solid pyrolysis time (Equation (2.15)) and the gas
induction time (Equation (2.17)), i.e.
ΛΓ−−
+−−−=
∞∞∞ oradrsPPgg
IPsssig xuqqTTxucc
TTct 1ln
/41
])()/([4)(
2""2/1
2
ρλρπλ
(2.21)
More accurate analyses can be developed for other complicated situations by numerically
solving both the solid and gas phase governing equations. For example, Wang and Yang
(1992) gave such example of the ignition of a polymethylmethacrylate (PMMA) sphere in a
forced oxidizer flow.
2.4.1.4 High surface temperature ignition
When a solid fuel slab is heated by an external radiation, the solid surface temperature may
reach a high level enough to cause gas phase ignition. This kind of ignition normally occurs
best with charring materials because the char layer generated by the initial pyrolysis can
reach a very high temperature due to radiation absorption or char oxidation. The high
temperature char in fact acts as the ignition source. This ignition process may take place in
the initiation of wood fire, especially when a sudden enhancement of solid gasification
happens.
The ignition model discussed in the foregoing subsections can be applied to describe this
kind of ignition, although the formation of char will affect the solid heating process and
some modifications are needed. If the ignition occurs before charring becomes important
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
41
(high surface heat flux, low charring materials), the ignition times are short and can be
described well by Equation (2.21). Here, ambient temperature T∞, i.e. the temperature of the
oxidizer flow, will be substituted by the surface temperature of the solid Ts in the induction
term, since the surface temperature (char temperature) may be higher than the temperature
of the hot gas flow. However, complicated phenomena such as transit from the char surface
to flaming are still not understood clearly and are thus impossible to model (Fernandez-
Pello 1995).
Some issues related to this problem have been studied. Alvares and Martin (1971) studied
how gas properties influence the solid ignition temperature and ignition time. Their research
showed that the surface temperature at the time of ignition and the ignition time decrease as
the ambient oxygen concentration or the pressure increase. Meanwhile, Ohlemiller and
Summerfield (1971) studied in-depth absorption. Kashiwagi (1985) studied the surface
oxidation. All these results are helpful to extend the above model and may need to be taken
into consideration while modelling heavy charring materials.
2.4.2 Piloted ignition
The piloted ignition phenomenon is similar to the previous problem but there is an extra
pilot source present. The pilot source can be a spark plug, a high temperature surface or a
flame. The function of this pilot source is to provide sufficient energy to ensure the gas
phase chemical reaction reaching ignition.
This ignition problem has been studied by Quintiere and co-workers (Quintiere 1981;
Quintiere and Harkleroad 1984). The results for a vertical solid slab heated by a radiative
flux are provided as follows. When the ignition is achieved with a pilot flame, the gas
induction time is small and the ignition time is very close to the pyrolysis time. The solution
of the problem is given by Equation (2.12) where the pyrolysis time can be replaced by the
ignition time and the pyrolysis temperature by the ignition temperature.
2"
2
)]([4
)(
Iigrad
Iigsssig TThq
TTct
−−−
=ρπλ
(2.22)
where:
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
42
h: heat transfer coefficient incorporating the convective and linearized radiation heat
transfer, and
)(, ∞−=+= TTChhhh sradradconv
hconv: convective heat transfer coefficient, kW/m2⋅K;
C: coefficient for linearized radiative heat transfer, kW/m2;
sT : average surface temperature, K.
The solid initial temperature has been taken as equal to the ambient temperature and the
surface temperature as equal to the pyrolysis temperature.
The minimum heat flux for solid ignition, min,igq ′′ , is the heat flux for the solid surface
coming into thermal equilibrium. It can be determined by applying ∞→igt :
)("min, Iigigig TThq −= (2.23)
where hig is heat transfer coefficient at ignition, kW/m2⋅K.
By investigating the pilot flame ignition of a vertical thick solid exposed to a constant
external radiation, Quintiere and Harkleroad gave descriptions of the relationship between
pilot ignition and flame spread (Quintiere and Harkleroad 1984). Based on Equation (2.23),
they developed a test method, LIFT (Quintiere 1981), for determining one of the
flammability characteristics, ignition time. The other critical parameter in this type of
ignition is the pyrolysis or ignition temperature. This value can be measured directly by a
thin type of thermocouple embedded at the solid surface, or determined experimentally by
obtaining the critical heat flux for the ignition, described in Equation (2.23). This critical
heat flux can be a radiative flux or a combination of radiative and convective heat fluxes.
Details of this method will be discussed later in this chapter.
2.4.3 Ignition criteria
Following a wide study of research results from various researchers, Kanury (2002)
summarized the ignition criteria for a solid as follows:
1. Ts ≥ T1,cr-- critical surface temperatures (Simms 1963; Martin 1965)
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
43
2. −T s ≥ T2,cr--critical average solid temperature (Simms 1963; Martin 1965)
3. crpy Rm ,1" ≥ --critical pyrolysis mass flux (Bamford et al. 1946)
4. δc≥δc,cr--critical char depth (Sauer 1956)
5. crg RtT ,2/ ≥∂∂ --critical local gas temperature increase rate
6. cri
i Rdyk ,3≥∞
--critical total reaction rate in the boundary layer (Kashiwagi 1974)
where:
sT and sT : surface and average surface temperatures, K;
T1,cr and T2,cr: suggested critical temperatures, K;
pym ′′ : pyrolysis mass flux, kg/m2⋅s;
R1,cr: suggested critical pyrolysis mass flux, kg/m2⋅s;
cδ and crc,δ : char depth and suggested critical char depth, m;
tTg ∂∂ / : local gas temperature increase rate, K/s;
R2,cr: suggested critical local gas temperature increase rate, K/s;
dyki
i∞
: total reaction rate in the boundary layer, 1/s;
R3,cr: suggested critical total reaction rate, 1/s.
Among these parameters, some are difficult to measure, like the critical gas temperature
increase rate and critical total reaction rate (relying on local fuel concentration
measurement); some are dependent on other parameters, like the char depth, which is
determined by surface temperature. The widely adopted criteria are critical surface
temperature and critical pyrolysis mass flux. They have been used in a number of ignition
models, and are discussed below. The possibility of applying some gas phase criteria is also
investigated in this section.
2.4.3.1 Critical surface temperature
Kanury (1995) suggested that the possibility of an ignition occurring under a given set of
conditions is judged by whether the exposed surface will attain a critical ignition
temperature. The temperature of the surface at the time of ignition is defined as the ignition
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
44
temperature, Tig. He also gave an approximate Tig for a broad range of timbers in a small
specimen tested in the vertical orientation as follows:
For spontaneous ignition: 600 oC for radiation exposure,
500 oC for convective exposure;
For piloted ignition: 300 ~ 410 oC for radiation exposure,
450 oC for convective exposure.
He also indicated that in spite of significant physical and chemical differences in structure
and composition, most organic solids undergo pyrolysis in a rather narrow temperature
range of 325±50 oC or 598±50 K.
By studying the ignition of radiantly heated wood in the presence of a small pilot flame,
Koohyar (1968) gave the critical temperature of about 600 K, near the pyrolysis temperature.
Alvares (1964) indicated that the ignition temperature increases with decreasing irradiance.
Recently, Babrauskas carried out an extensive literature review of the ignition temperatures
of timber (Babrauskas 2001). He summarized that piloted ignition at heat fluxes sufficient to
cause a direct-flaming ignition normally occurs at surface temperatures of 300 ~ 365 oC
(300 ~ 310 oC for hardwoods, and 350 ~ 365 oC for softwoods). The ignition temperature is
around 250 oC for wood exposed to the minimum heat flux, which mostly causes a glowing
or glowing/flaming combustion. The results for auto-ignition temperatures evidently span a
huge range, from 200 to 700 oC while the median values are from 380 to 500 oC.
Babrauskas suggested that following reasons should be considered in relation to these
widely differing ignition temperatures:
The definition of ignition used
Piloted vs auto-ignition conditions
The design of the test apparatus and its operating conditions
Specimen conditions (e.g., size, moisture, and orientation)
The species of wood
It can be seen that it is difficult to find a widely acceptable and applicable ignition
temperature as the ignition criterion for a timber, as all these reasons contribute to wide data
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
45
variation. Even for tests carried out in a test apparatus, it is found that the ignition
temperature is dependent on a number of parameters. One of these is radiation flux, and the
gas phase condition over the solid surface is another. This gas phase condition includes gas
flow velocity, oxygen concentration and ambient pressure (Fernandez-Pello 1977;
Kashiwagi and Omori 1988; Blasi et al. 1989). All these factors must be investigated and
evaluated where the ignition temperature is considered as the ignition criterion.
2.4.3.2 Critical mass flux
Compared with the great amount of ignition temperature data, experimental results for
critical mass flux are relatively scarce. However, the critical mass flux is suggested by a
number of researchers to range from 0.8 to 6 g/m2⋅s for polymers (Tewarson 1982; Deepak
and Drysdale 1983; Rasbash et al. 1986; Thomson and Drysdale 1989; Tewarson et al.
1999). An example value that Banford (1946) obtained for wood is 2.5 g/m2⋅s.
It is noticed that the critical mass flux changes when radiative flux and gas flow conditions
changes. For example, Yang et al. (2003) reported 2 to 4 g/m2⋅s for the critical mass flux for
cherry wood and the value of this mass flux increased as the radiative flux increased.
Another example of varying gas flow conditions is the experimental study of PMMA carried
out by Zhou et al. (2002). These tests were conducted in the Forced-flow Ignition and
flame-Spread Test (FIST) apparatus, which was developed at the University of California,
Berkeley (Cordova et al. 2001). This apparatus is a derivation of the LIFT apparatus. The
gas flow was forced over the sample surface which was heated by a radiative panel. A spark
plug was placed 0.5 to 1.5 mm above the PMMA plate. In the tests the researchers found
that the critical mass flux increased from 1.5 to 3.0 g/m2⋅s when the laminar parallel air flow
over the surface increased from 1 to 1.7 m/s.
Alvare’ s research conclusion, which was quoted by Kanury (2002), indicates that the
parameters discussed above, like critical temperature, pyrolysis flux and total gas reaction,
are not generally acceptable as criteria of ignition. This is because these parameters are all
strongly dependent on other factors, such as exposed heat flux, boundary layer height and
slab thickness, re-radiative loss and chemical kinetics. As there is not enough detailed
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
46
information about these factors, or suitable and effective computation available, they are
questionable for adopting as criteria.
2.4.3.3 Lower flammable limit and possible criteria in gas phase
For many years the concept of combustion has been symbolized by the Triangle of
Combustion and represented as fuel, heat and oxygen. Further fire research determined that
a fourth element, an uninhibited chemical chain reaction, was a necessary component of
combustion. Then the fire triangle was changed to a fire tetrahedron (a pyramid) to reflect
this fourth element (Haessler 1974), as shown in Figure 2.3. It is know that an uninhibited
chemical chain reaction only occurs within certain reactant concentrations. In combustion,
these suitable reactant concentrations in gas phase are normally expressed as within a range
that was confined by the lower and upper flammable limits. Therefore the fourth element for
the combustion tetrahedron was also suggested as an item “ Atmosphere” (suitable fuel
concentration in a favoured environment).
Figure 2.3 Combustion tetrahedron (pyramid)
Some other gas phase parameters, which were based on the concept of above tetrahedron,
have also been considered as the criteria of ignition by researchers (Wichman 1986;
Kashiwagi et al. 1990; Durbetaki et al. 1995). Through a verification of these criteria
through gas phase ignition experiments, Tzeng et al. (1990) found that the fuel
concentration at the location of the ignition source and at the time of ignition for all gas
phase ignition cases in their study was approximately the same. Zhou and his co-workers
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
47
(2002) obtained similar results through both experiments and numerical analysis for ignition
of polymer materials. They found that the gas fuel concentration above the solid, which is
partially determined by the pyrolysis rate, is more meaningful compared with the critical
mass flux. One conclusion that the authors reached is:
“ The invariant gas mixture composition at the time of ignition indicates that
for ignition to occur a particular stoichiometry must be reached at the pilot
location. This critical stoichiometry, most likely related to the lean
flammability limit, is obviously a good choice as the ignition criteria.”
One widely adopted concept for assessing the lean flammability limit is the so-called lower
flammable limit (LFL). This is the limit of composition within a fuel/diluent mixture which
a flame can propagate. It is expressed as a concentration of fuel in a specified
oxidant/diluent mixture at a specified temperature and pressure.
Another widely used concept is the limiting oxygen index (LOI) for supporting flame
propagation. This is normally defined as the oxygen mole fraction of the oxidant stream at
the point of flammability. The LOI can be used to predict the extinction of flame under
certain diluted oxidant conditions. Compared to the LOI, the LFL is more useful in
evaluating the possibility of ignition in the fuel mixture.
As indicated by Beyler (2002), the ability of a fuel and oxidant pair to react in diffusion
flames can be evaluated by examining the flammability of a premixed stoichiometric
mixture of the fuel and oxidant. This may be achieved by satisfying Le Chatelier’ s empirical
rule. Le Chatelier’ s rule, which was generalized by Coward et al. (1919), is expressed as
follows in an ambient atmosphere:
=≥
n
i i
i
LFLC
1
1 (2.24)
where:
Ci: volume percent of fuel gas, i, in the fuel/air mixture;
LFLi: volume percent of fuel gas, i, at its lower flammable limit in air alone.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
48
If the indicated sum is greater than unity, the mixture is above the lower flammability limit.
This relationship can be restated in terms of the lower flammable limit concentration of the
fuel mixture, LFLF, as follows:
( )=
= n
iifi
F
LFLCLFL
1
/
100
(2.25)
where Cfi is the volume percent of fuel gas i in the fuel gas mixture.
The lower flammability limit at other temperature conditions can be obtained from the
above values by applying the critical adiabatic flame temperature at the lower flammable
limit. The concept of the critical adiabatic flame temperature came from observation by
Burgess and Wheeler (1911). They noted a constancy of the potential heat release rate per
unit volume of normal alkane/air lower flammable mixture at room temperature. This
observation also implies that the adiabatic flame temperature at the lower flammable limit is
a constant, since the heat capacity of the products from complete combustion are almost the
same for all hydrocarbons. The details about this concept and its utility exceeds the research
area of current study, but can be found in Beyler’ s paper (Beyler and Hirschler 2002).
Due to the constancy of the adiabatic flame temperature at the lower limit, this concept can
be used to predict the effect of variable mixture temperature and diluents. The relation
between the LFL and the critical adiabatic flame temperature, Tf,LFL, is given by the
following equation when all the heat released in adiabatic combustion is absorbed by the
products of combustion:
=∆
LFLf
I
T
T pc dTnCHLFL ,
100
(2.26)
where:
∆Hc: heat of combustion of the fuel, kJ/kg;
LFL/100: mole fraction of fuel;
n: number of moles of products of combustion per mol of fuel/air mixture;
Cp: heat capacity of the products of combustion, kJ/kg⋅K;
TI: initial temperature of the fuel/air mixture, K;
Tf,LFL: adiabatic flame temperature of a lower flammable limit mixture, K.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
49
The heat capacity in Equation (2.26) is a temperature-dependent variable. When an average
value of the heat capacity is applied in the above equation, it gives:
( )ILFLfpc TTnCHLFL −=∆
,100
(2.27)
Equation (2.27) exists for different initial temperatures. Therefore the following equation
applies for LFLs at various initial mixing temperatures.
( ) ( )100/100/ 2
2,,
1
1,,
LFL
TT
nCH
LFL
TT ILFLf
p
cILFLf −=∆=
−
(2.28)
Suppose the first initial temperature TI,1 is the ambient temperature, then the LFL at another
initial temperature TI,2 can be calculated from Equation (2.28).
It is postulated that flammability limit is a more appropriate criterion for ignition.
Flammability limits are expressed in terms of fuel volume fraction in air. Since mixture
fraction is defined as fuel mass fraction and, therefore, fuel volume fraction has one-to-one
relationship for any given fuel. The lower flammable limit can be translated to a critical
mixture fraction, which is a form of critical gas phase fuel concentration.
The concept and theory of the LFL are simple and the relationship between the LFL and
critical mixture fraction is clear. An ideal approach to validate this criterion is to compare a
LFL value from the theoretical computation with a fuel concentration value (expressed as
the mixture fraction) measured from experiments. However, the direct application of this
concept is not so easy. The reason is that the experimental measurement for the mixture
fraction of pyrolysis products is restrained by current technology and resources. Therefore
investigation of the mixture fraction as the criterion will rely on some type of numerical
computations. Such numerical solutions may require advanced computing technology and
the availability of more accurate decomposition chemistry, gas phase chemical kinetics and
thermal and radiative properties (Zhou et al. 2002). This reveals a possible research
approach for this study to achieve an applicable criterion, some form of fuel concentration,
for complex gas phase conditions.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
50
2.5 Flame Spread on Solid Combustible Surfaces
As mentioned in Chapter 1, flame spread over solids is a key component in a fire model
since this will affect the initial fire development and heat release. This has been a subject of
interest in both testing and modelling researches.
2.5.1 Flame spread mechanism
Flame spread over a solid combustible surface can be regarded as a continuous pilot ignition
process. For the flame to spread, enough heat must be transferred from the flame and/or
other heat sources to the unburnt area ahead of the flame front. The vaporized fuel is then
diffused and convected away from the surface to mix with the oxidizer. The flammable
mixture is formed ahead of the flame edge and is then ignited by the flame.
Flame spread was reviewed in detail by Williams (1976). The fundamental equation for the
flame spread is as follows:
"qhV f =∆ρ (2.29)
and
)( Iig TTch −=∆ (2.30)
where:
: density of the medium, kg/m3;
Vf: flame spread speed, m/s;
"q : net heat transfer, kW/m2;
c: specific heat of the solid or liquid, kJ/kg⋅K,
Tig: ignition temperature, K; and
TI: initial surface temperature, K.
Rearranging Equations (2.29) and (2.30):
( )Iig
f TTcq
V−′′
=ρ
(2.31)
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
51
The problem is essentially about heat transfer. The geometry of the solid, i.e. horizontal or
vertical orientation, opposed or concurrent flow, is found to affect the heat transfer, and has
been studied extensively (Quintiere 1995).
The heat transfer from the flame to the unburnt combustible ahead is strongly dependent on
the orientation of the flame, which in turn is dependent on the characteristics of the gas flow.
When the gas flow, either naturally induced or forced, opposes the direction of flame spread,
the flow keeps the flame close to the surface downstream of the pyrolysis front. This type of
spread is commonly called opposed flow flame spread and natural convection horizontal
flame spread is an example. If the gas flow is in the direction of spread, the flame is pushed
forward ahead of the pyrolysis region. This type of spread is known as concurrent flame
spread and natural convection upward spread of flame is such an example.
Fernandez-Pello et al. (1980) studied both opposed flow and concurrent flow flame spreads,
and found a similarity between the mechanisms controlling the solid ignition and the flame
spread processes. Their experimental results show a relationship between the ignition delay
time and the flame spread rate. In the flame spread process, the flame acts as both a heating
source for pyrolysis and a pilot for ignition. This suggests that a simplified model of flame
spread could be based on the analyses developed for solid ignition (Section 2.4). Since the
time for a solid element to ignite is the same as for the flame to propagate to the solid
element position, the velocity of flame spread will be given by the ratio of the solid heated
length ahead of the pyrolysis front to the solid ignition time. For a comprehensive analysis
of flame spread under various conditions, readers are referred to papers by Fernandez-Pello
(1995) and Quintiere (1995).
Theories of flame spread have been built up, and a number of models to apply the theories
have been published (Drysdale 1999).
Quintiere and Rhodes (1994) developed a model for the combustion of PMMA in the cone
calorimeter. The analytical model was based on solutions for one-dimensional unsteady
conduction and surface vaporization at a fixed temperature. The burning rate was calculated
by the moving vaporizing surface at the fixed vaporization temperature. The flame radiation
was calculated by adopting an algorithm for pool fires to obtain the absorption coefficient
and mean beam length. This is a relatively simple model, and suitable for plastic material,
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
52
such as PMMA, since their vaporization and combustion behaviour is similar to that of a
pool fire.
Fernando (2000) developed a CFD-type flame spread model, CESARE-CFD, that was based
on a simple combustion model which was combined with a flame spread model (Luo and
Beck 1996). This CFD model adopted the basic ignition controlling mechanism described in
Section 2.4 and a critical surface temperature was used as the ignition criterion. It used the
kinetics obtained from TGA tests to model the pyrolysis of polyurethane (PU) foam with
certain assumptions about the decomposition process. The simulated prediction matched
favorably with tested flame spread on a PU foam slab in a full-scale room test.
Flame spread on building elements, especially on walls, is an important application area for
basic flame spread theory and the material property data obtained from tests. Quintiere
suggested a practical approach to predicting material performance (Quintiere 1988). In his
study, an analytical method was adopted to create correlations between the theories of
ignition, spread and combustion and the tested material data. These properties have been
measured for many types of materials, such as thermo-plastics and timber.
Beyler et al. (1997) developed a model to describing upward flame spread and fire growth
on wall materials. This model includes following sub-models: ignition, material heating,
pyrolysis and burning rate, flame spread, and flame and surface heat transfer. In this model,
a relatively simple pyrolysis model was adopted which determining the pyrolysis flux by
effective heat flux and heat of gasification of the material. The heat transfer was simplified
as one-dimensional problem and integral solution was used. The ignition criterion is the
critical surface temperature. The material property data, including the thermal inertia, kρc,
ignition temperature, effective heat of gasification, and heat of combustion, were obtained
from cone calorimeter tests. This model has been used to predict flame spread behaviour of
PMMA, plywood, and vinyl-ester/glass composite, and reasonable agreement with
experimental data has been achieved.
Generally flame spread depends on the heat transfer processes at the front of the flame.
These transport processes depend on both the fuel material properties and other
environmental conditions. Therefore the estimation of flame spread requires specific
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
53
material data and complex analysis. The current state of knowledge provides limited
formulas and material data to make this estimation.
2.5.2 Material properties and measurement for the flame spread
Since flame spread is a continuous ignition process, the two mechanisms that control
ignition, discussed in Section 2.4, can also be assumed to control flame spread. The material
properties that are associated with the controlling mechanisms include density, ρ, thermal
conductivity, k, specific heat, c, and ignition temperature Tig. These parameters are
sometimes treated as parameter groups under specific circumstances. For example, when
dealing with thermally thick materials, the predominant parameters are kρc and Tig. If the
material studied is thermally thin, then parameter group ρc and Tig are important.
For engineering applications of thermal modelling, it is important to develop a practical
methodology to estimate these properties. Tig is a critical parameter in the application of
both the ignition and flame spread models. As noted earlier, its value can be measured
directly with thin thermocouples embedded at the surface of the solid. However, the
measurement may not be very accurate because it is difficult to place the thermocouple
exactly at the surface, and it may also absorb or emit radiation in a different way from the
solid surface itself. Therefore some analytical methods, combined with the experimental
data to estimate this parameter, are still necessary.
2.6 Experimental Methods
Experiments or tests play a key role in fire science. Fire may be the first and also the most
important tool in history of human beings. It has been utilized for more than thousands of
years. However, studies of building fire behaviour, especially the modelling of fires, are still
in their early stages. All fire models available for fire engineering users need experimental
data or results either as inputs or for validation purposes. First, experiments can introduce
concepts and semi-empirical formulas, such as the “ zone” concept. Secondly, they also
provide necessary data, such as material properties. Thirdly, a model can only be put to
practical purposes after being validated by experiments.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
54
Experiments used in fire science can be divided into two categories: bench-scale tests and
full-scale tests. Bench-scale tests can provide material properties and their basic behaviour
in fires. They are relatively cheap and can be repeated easily. The commonly used small and
bench-scale experimental methods and apparatus include Thermogravimetric analysis
(TGA), the cone calorimeter, Lateral Ignition and Flame Transport (LIFT), and Single
Burning Item (SBI) tests.
On the other hand, middle to full-scale tests can simulate real fire scenarios, measure fire
behaviour (mainly the heat release rate), and validate the modelling results. However, they
may be expensive. The middle to full-scale test equipment includes the furniture calorimeter
(NT FIRE 032), the room calorimeter (ISO 9705), and the room/corner test device (ASTM
1983). The furniture calorimeter enables the study of combustion behaviour of a whole
piece of furniture and one example of application of this equipment can be found in research
carried out by Babrauskas et al. (1982). One example of the full-scale tests is a doorway
calorimetry that can be used to measure oxygen concentration profiles across the burn room
door (He et al. 1998). Another such example is the room/corner test that is used to evaluate
the fire performance of wall and ceiling linings from a corner fire. This test method has been
standardized by ASTM (ASTM 1983) and ISO (ISO_9705 1993).
The middle to full-scale experimental methods will not be reviewed individually in this
literature review since they are not directly related to the current research interest.
2.6.1 Thermogravimetric analysis (TGA)
The thermogravimetric (TG) analyser is now a standard device in chemistry and chemical
engineering research. It was first used for analysis of coal burning profiles by Wagoner and
Duzy (1976). This test has been established as a routine procedure and is in regular use for
assessing the burning characteristics of solid fuel, such as coal products. More recently it
has been used to study reaction kinetics (rates) of pyrolysis and combustion (Stanmore
1991).
The TG analyser maintains a desirable thermal condition for a specimen of a very small
quantity (less than 20 mg) and a constant supply of a relatively large amount of working gas
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
55
so that the diffusion control factor of the reaction can be neglected and hence the reaction
kinetics can be analysed. The output from a thermo-balance system is plotted against
temperature to obtain the kinetic features of the sample. A detailed description of this device
is given in Section 4.1.
Research using the TGA to calculate the kinetics of materials can be found in works by
Ramiah (1970), Cordero et al. (1989; 1991), Parker and LeVan (1989), and Liu (2000).
Questions have been raised about the direct application of the kinetics obtained from the
TGA tests (Schneider 1992; Beyler and Hirschler 2002). Concerns about the traditional
TGA method are the low heating rate and large temperature gradients in the specimen. The
heating condition monitored is limited to the surface or a shallow depth of the specimen.
This may limit the application of the kinetics in flaming combustion conditions.
To eliminate the above problems and obtain applicable kinetics, a number of efforts have
been made. Cordero et al. (1989) suggested controlling the sample size and heating rate for
the TGA tests. Parker (1988) developed a test method to separate the whole set of apparatus
into two parts, the TG analyzer (as pyrolyzer) and the catalytic converter (carbon dioxide,
water vapor and oxygen analyzers). This was believed to help eliminate the effect from the
heating environment (in the pyrolyzer) for the calculation for the kinetics (through the gas
analysers). On the other hand, improvements have been also made in the computational
scheme for the kinetics. Seungdo and Park (1995) demonstrated the application of high
heating rates in TGA tests. They also verified a shifting pattern in reaction rate curves,
which was discovered by Ozawa et al (1970) and is capable for calculation of the kinetics.
Liu and his co-workers (Liu and Fan 1998; Liu et al. 2003) improved the kinetics
calculation by illustrating a kinetic compensation effect in the kinetic parameters and
developing a method to avoid the presupposition for reaction order, which may introduce
errors. Staggs’ research (1999) also demonstrated that a global in-depth pyrolysis model
using TGA-derived kinetics may be able to predict mass loss rates in bench-scale
experiments such as the cone calorimeter tests.
Despite these improvements, a key issue still somehow remains. As discussed in Chapter 1,
it is a major challenge to find a relatively simple test method and computational scheme to
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
56
obtain “ effective” kinetics that can adequately describe the decomposition process under
real fire conditions.
2.6.2 Cone calorimeter
The cone calorimeter is the most commonly used bench-scale heat release rate apparatus. It
measures the heat release rate (HRR) by utilizing the oxygen consumption principle. It was
developed by Babrauskas at the National Bureau of Standards in the 1980s (Babrauskas
1984).
The cone calorimeter simulates the conditions of a piece of fuel exposed to a fire
environment. The major measurements include the mass loss rate (MLR), heat release rate
(HRR), temperature, species yield, radiation, and so on. The apparatus and test procedure
are standardized in the USA (ASTM_E1354-04 2004) and internationally (ISO_5660-1
2002).
The cone calorimeter can be used to measure certain quantities in the testing conditions.
These quantities include heat release rate, ignition time, heat of combustion, etc. These
measured quantities are used to predict fire behaviour in a full-scale or compartment
environment. Detailed information about this bench-scale experimental device and its
operation is included in Chapter 5. There is a great wealth of literature in regard to the cone
calorimeter and its applications in fire research. Only a few articles that clearly relate to the
current research interest are reviewed below.
Grexa and co-workers carried out a series cone calorimeter tests on wood products (Grexa et
al. 1996; Grexa et al. 1997). The major measurements in their tests were HRR, total heat
released (THR), and ignition time (tig). A number of thermal properties were then derived
from those basic measurements, such as critical heat flux "crq by Janssens’ method
(Janssens 1991), thermal inertia kc and heat of ignition ∆Hig. The effects of wood species
and timber surface orientation (alone or cross-sections) in the cone calorimeter tests were
investigated by Harada (2001). While most of experimental data available were obtained
from the along-texture section samples, a slower combustion (longer tig and smaller HRR at
the same external radiation level) was observed for the cross-texture section samples. Tests
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
57
for the thermal plastic material PMMA were carried out by Tsai et al. (2001) and Zhou et al.
(2002). It was found that ignition process depends not only on PMMA pyrolysis but also on
gas phase reaction (i.e. it is affected by gas flow conditions, which matches previous
analysis for ignition). As a thermal thick material, a linear relationship between igt/1 and
radq ′′ was observed from the tests for the PMMA.
By running Round-Robin type international or inter-laboratory tests (ASTM 1990), errors in
the material property measurement from the cone calorimeter, which are associated with
device and operation differences, can be reduced greatly.
Meanwhile the combustion process in the cone calorimeter has been modelled by a number
of researchers (Staggs and Whiteley 1999; Bilbao et al. 2001; Janssens et al. 2001). Some
reasonable simulation results have been obtained. In general, the standardized testing
condition and environment makes the cone calorimeter a useful device for measuring
material properties and to validate ignition models.
More cone calorimeter experimental results, especially for those materials identical or
similar to the furnishing materials used for this research will be discussed later in Chapter 5.
2.6.3 LIFT and SBI
LIFT (ASTM_E1321-90 1990) is a flame spread test resulting from Quintiere’ s research
(Quintiere 1981). This device has been used to measure the “ effective” properties for
ignition and flame spread, including ckρ and Tig. Measurements from LIFT are normally
combined with other bench-scale test results, such as those from the cone calorimeter, to
predict full-scale and real fire environment behaviour. Some applications of results from this
testing can be found in the literature (Quintiere and Harkleroad 1984; Delichatsios 1999;
Zhou et al. 2002).
The SBI test (prEN_13823 2002) is a new fire test method developed for the Euroclass
system. The test is based on a fire scenario of a single burning item located in a corner
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
58
between two walls covered with the lining material to be tested. The SBI test can be used
for the classification of construction products.
Measurements from the SBI test include the heat release rate and the smoke production rate.
Observation of flame spread can also be made, but only to the extent of recording the time
taken to reach the extreme edge of the sample. These data are used to produce derivative
indices, such as FIGRA (Fire Index Growth RAte) and SMOGRA (SMOke index Growth
RAte), for classification. These two indices express the rates of growth of the heat release
rate and the smoke production rate.
Modelling and experimental studies have been carried out for the relationship between the
SBI test and the cone calorimeter test (Messerschmidt et al. 1999; Hakkarainen and Kokkala
2001). It was found that by using cone calorimeter tests, the number of SBI tests required
could be reduced greatly without additional data or material parameters. For example, the
results from a single cone calorimeter test are useful for predicting peak heat release rates in
a number of SBI tests (Messerschmidt et al. 1999).
2.6.4 Applications of the experimental methods
This section describes the methods that may be used or have been developed to obtain
several thermal parameters, by the above experimental methods.
The “ effective material fire properties” for modelling flame spread can also be measured by
a number of test procedures. The lateral ignition and flame spread (LIFT) apparatus is one
such example (ASTM_E1321-90 1990). The effective properties obtained from this method
include: ckρ , Tig and (a measurement of the flame heat transfer under conditions of
opposed flow natural convection in air).
The ignition temperature can also be measured in cone calorimeter tests following an
experimental procedure suggested by Quintiere and Harkleroad (1984). A series of cone
calorimeter tests were carried out to determine the minimum radiative flux required for a
kind of fuel ignited. The minimum radiative flux is decided by the radiative flux when an
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
59
asymptotic curve of ignition time is extended to the infinite (normally 600 seconds is used).
The experimental results are illustrated in Figure 2.4. Then Equation (2.23) is applied to
calculate the ignition temperature when the heat transfer coefficient at ignition hig is known
or can be determined from other methods.
Figure 2.4 Experiments for obtaining ignition temperature
(quoted from work of Fernandez-Pello (1995))
The heat of combustion, ∆Hc, can be determined from the cone calorimeter test (using
measured heat release rate and a mass loss rate) by a procedure suggested by Janssens
(1992). The calculation of effective heat flux for a thermally thick solid was modified by
extensive numerical simulation results.
There are other parameters not related to the current study and thus not discussed in this
thesis. Table 2.1 below briefly summaries the major properties of solid combustible
materials and the measuring devices.
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
60
Table 2.1 Major parameters and measuring methods
Parameters Measuring apparatuses
Ignition temperature, Tig cone calorimeter or LIFT
Thermal inertia, kρc as above
Lateral flame spread parameter LIFT
Minimum temperature for lateral spread, Ts,min as above
Heat of combustion, ∆Hc cone or bomb calorimeter
Effective heat of gasification, L cone calorimeter
Heat release rate per unit area, Q’ ’ as above
Activation energy, E, & pre-exponential factor, A TG analyser
However, it should be noted that the capability of a bench-scale test apparatus to reproduce
end-use conditions of tested materials and correspondence of test conditions with fire
environment is not yet very clear. Originally the heat release rate of common combustible
building and furniture materials was believed to be sensitive to the heating condition. It is
technically difficult to run cone calorimeter tests at very weak levels of radiation (say, less
than 10 kW/m2 for timber materials) which is still meaningful in flame spread for fire safety.
Meanwhile, the heat release rate obtained from the cone calorimeter at a level of external
heat flux of 25 kW/m2 or lower is considerably lower than that obtained at an identical
heating condition in full-scale or medium-scale heat release measurement. While the bench-
scale test results were difficult to apply to such situations, basic pyrolysis analysis (based on
the kinetics from TGA tests) and heat transfer calculations may be able to contribute to the
solution of this kind of problem.
2.7 Unresolved Phenomena and Research Interests
Unresolved issues based on the literature review and discussion in the previous sections
were summarized in this section. Interested areas and requirements were then identified to
achieve the specified research aims described in Chapter 1.
2.7.1 Unresolved issues related to current research
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
61
Providing enough accurate information to describe fire development is necessary in
performance-based fire safety and risk assessment practice. It is also a requirement for
development of today’ s fire models. This task requires a better understanding of fire itself.
For example, the heat release rate is the single most important variable in describing fire
hazard (Babrauskas and Peacock 1992). The prediction of HRR has experienced continuous
improvement during the last decades as fire models have developed. As reviewed previously,
this improvement started from prescribing input data for the HRR, to perform basic
calculations for simple fuel configuration, such as pool fuel, wood crib or polyurethane
foam, and then estimating real-scale HRR based on bench-scale test results. The limitation
of this latter scheme is numbers of uncertain and then assumptions existed when applying
the bench-scale HRR values to the prediction of real-scale HRR.
Recently prediction of HRR based on a combustion model, which combined with
computation from a flame spread model, has been adopted into a number of fire models
(Drysdale 1999). However, serious limitations still exist for applying the theories into real
flame spread prediction as a number of assumptions and empirical correlations have to
embedded into the models. Examples include the NRCC-VUT (He 1996) and FIRST (Mitler
and Rockett 1987), etc. Even for a CFD model, which has the ability to describe the
combustion process in detail, the flame spread modelling is still relatively crude. There are
number of reasons. First, most CFD models have only one-dimensional heat transfer ability
for the solid (Quintiere and Rhodes 1994; McGrattan et al. 2000). A number of simplifying
approximations for the geometry and heat transfer conditions are normally made to make a
flame spreading problem theoretically ‘tractable’ . This often limits the usefulness of the
models. Secondly, modelling of flame spread on a solid greatly depends on ignition
properties of the material, such as ignition temperature. Such critical data are still very
limited for widely used materials (Babrauskas 2001). Finally heat transfer in real fire
conditions is very complicated and very fine resolutions of grid in CFD models are required.
This limits the engineering applications of most of today’ s computer models. Reasonable
accuracy can only be achieved by modelling more details about the combustion phenomena
with a robust numerical scheme.
Fire model developers are currently making improvements for numerical schemes aiming to
deal with complicated heat and mass transport, and to shorten the simulation duration. Some
progresses have been achieved recently. Therefore the research task of this study returns to
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
62
the area identified in Chapter 1: the basic phenomena within the pyrolysis, ignition and
flame spread processes.
The basic theories and experimental methods for measuring quantities related to these
processes have been developed as reviewed in previous sections. There are still unresolved
issues, which have been identified in this literature review related to the current research
aims. These can be listed as follows and some are similar to those summarized by T’ ien
(2002):
1. Lack of detailed chemical kinetic information for the combustion of solids. This
includes both the solid thermal decomposition processes and the oxidizing
kinetics of the pyrolysis gases.
2. The application to a real fire condition of the kinetics obtained from the normal
TGA test method and computational scheme is still questionable.
3. Solid phase processes, for the studied materials, may include charring, bubbling,
cracking and smouldering. Whether these events should be incorporated into the
model may depend on their importance to a particular problem.
4. The common ignition criteria for solid materials are still to be validated in
complicated geometry conditions. The criterion associated with gas phase fuel
either does not exist or is difficult to compute.
Meanwhile, lack of knowledge of widely used materials, even for commonly used
furnishing materials, impedes their applications in fire models. This is partly due to
insufficient applicable experimental data, and partly due to inadequate models (methods) to
interpret the available data. For example, the kinetics of the thermal decomposition
processes of the furnishing materials selected in the current study are either not known or
there is great variation in the available experimental data.
2.7.2 Research interests and requirements identified for current study
According to the unresolved issues identified above, the interests of the current research will
focus on pyrolysis analysis and ignition processes. The objective of the study is to verify the
critical mixture fraction as an ignition criterion, which will be applied in an engineering
approach for prediction of combustion and flame spread. The methodology is to use a
Chapter 2 Literature Review ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
63
combined experimental and numerical simulation approach. The better understanding of
pyrolysis progress enables describing pyrolysis and ignition in detail by suitable kinetics
and modelling. By applying those pyrolysis details a robust CFD fire model will be
improved to generate engineering satisfactory prediction of ignition based on the critical
mixture fraction criterion. The requirements for achieving this aim include the following:
Better understanding of the decomposition process of studied materials under
different environments
Improvement of the TGA test method and computational scheme to obtain the
“ effective” kinetics which are able to describe the decomposition process in real
fire environments
Modelling the pyrolysis process with major phenomena for the studied materials.
(Detailed requirements for the modelling will be discussed in the next chapter.)
Applying the above, modelling results into a CFD model to simulate pyrolysis,
ignition and combustion processes in a carried out bench-scale test, cone
calorimeter environment
Studying the ignition criteria by validating various possible parameters from the
modelling against the data from the experiments; calculating the gas phase fuel
concentration (the mixture fraction) via CFD modelling and obtaining the critical
value at the ignition time which determined by the bench-scale tests
Validating a relationship between the critical mixture fraction and the lower
flammable limit. Developing a gas phase criterion that based on the lean
flammability limit theory and applying it onto complex environmental conditions.
64
1234567!81234567!81234567!81234567!8 9;:<=6&>?>$@BADCE:F 526&79G3>9;:<=6&>?>$@BADCE:F 526&79G3>9;:<=6&>?>$@BADCE:F 526&79G3>9;:<=6&>?>$@BADCE:F 526&79G3><=61H:9I4J:K@%5@L:MAN3A<E@C<=61H:9I4J:K@%5@L:MAN3A<E@C<=61H:9I4J:K@%5@L:MAN3A<E@C<=61H:9I4J:K@%5@L:MAN3A<E@[email protected]@:A :[email protected]@:A :[email protected]@:A :[email protected]@:A :FKO:>$@B<E9G3'567&@P3Q>RKKO:>$@B<E9G3'567&@P3Q>RKKO:>$@B<E9G3'567&@P3Q>RKKO:>$@B<E9G3'567&@P3Q>RK
This chapter deals with modelling of decomposition process, including theoretical analysis
and numerical modelling. Based on analysis and comparison for available pyrolysis models,
one of the models that meets research requirements will be chosen as a prototype for further
modelling. This chapter consists of five sections. The first section describes pyrolysis
models in general. Criteria for choosing a pyrolysis model are developed. Section 3.2
discusses Atreya’ s pyrolysis model in detail. Basic equations for the heat transfer and
charring processes are presented. Section 3.3 introduces the major improvements for the
Atreya’ s model made by other researchers. A simple theoretical calculation is performed for
one of the studied materials, mountain ash, in Section 3.4. Section 3.5 gives a brief
summary of this chapter.
3.1 Pyrolysis and Ignition Models
As reviewed in Chapter 2, a substantial number of experimental and modelling researches
have been carried out to study the pyrolysis of a charring solid. The primary objective of
these experimental works has been identified as obtaining thermal quantities, such as
chemical kinetic data, heat of pyrolysis as well as pyrolysis mass fluxes. The primary
objective of pyrolysis modelling is to predict pyrolysis and ignition behaviours of both full-
scale and real fires based on measured thermal quantities.
3.1.1 Classification of pyrolysis models
During the past 30 years, a number of pyrolysis models have been developed (Moghtaderi
2001). These models, according to the technical basis adopted to describe the conversion of
virgin fuel into volatiles and char residues, can be classified into the following categories:
1. simple thermal models, including
algebraic and analytical models
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
65
integral models
2. comprehensive models, including
analytical models
numerical models
In the thermal models, the basic assumption is that pyrolysis occurs when the temperature
reaches a so-called “ pyrolysis temperature” . With this assumption the problem is greatly
simplified and only basic energy balance needs to be solved.
Algebraic and analytical thermal models ignore most of the chemical and physical processes
involved in order to reduce the complexity of the solution. For example the mass flux is
expressed as a function of location of the pyrolysis front (where surface temperature reaches
the pyrolysis temperature). The applicability of this type of model is limited due to the fact
that it is incapable of providing a comprehensive solution for detailed outputs. Models
developed by Mikkola (1991) and Kanury (1994) are a few examples of this type of models.
Integral thermal models employ numerical solution techniques to solve temperature
distribution and mass flux. It can take many physical phenomena into consideration.
However, like all thermal models, the integral model neglects chemical kinetics and is again
based on the critical temperature criterion. By reducing the partial differential equations
(PDE) for energy conservation into ordinary differential equations (ODE), the integral
thermal models are relatively simple, easy to use, and computationally economic. However
the accuracy for the results is not as high as that of PDE methods. Moghtaderi et al. (1997),
Spearpoint and Quintiere (2000; 2001) give examples of thermal integral models.
The critical pyrolysis temperature criterion used in the thermal models is equivalent to
assuming that chemical processes are much faster than diffusion processes. This has been
discussed in Section 2.4. Since the ignition temperature is assumed to be the pyrolysis
temperature, (i.e. tig = tpy), the gas phase induction time is ignored. However for common
solids the global pyrolysis is controlled by chemical kinetics at low temperatures and by
diffusion at high temperatures. A building fire may occur over a wide range of temperatures
(between 450 – 1000 oC), both diffusion and chemical kinetics should be taken into
consideration. This sets the underlying principle of comprehensive models.
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
66
In comprehensive models, chemical processes are normally based on first order kinetic
schemes. According to the schemes adopted, the models can be categorized as follows:
1. one-step global
2. one-stage multi-reaction
3. two-stage semi-global
The items “ step” or “ stage” deal with the number of stages involved for a virgin fuel to
decompose into products. For example, “ one-step” means that the conversion performs
directly without middle products being considered. The terms of “ global” or “ multi-
reaction” deal with how to treat the pyrolysis energies. For example the global scheme
calculates an activation energy to represent the combination effect of all sub-reactions,
whereas activation energies for a number of major reactions are calculated separately in the
multi-reaction. In a two-stage semi-global scheme, an activation energy is used for all sub-
reactions in one of the two stages.
An example of the development of the comprehensive model may be found in an improved
version of Kung’ s model (Kung 1972). Kung’ s original model incorporated features like
variable thermo-physical properties and convective heat transfer. Kung’ s model was further
developed to account for the porous structure of some solids (DiBlasi 1993), structure
changes (Parker 1985) and effect of moisture content (Atreya 1984; Moghtaderi et al. 1998).
According to a review given by Moghtaderi (2001), the typical kinetics for one-step global
pyrolysis of cellulose have been reported by many researchers. The activation energy ranges
from 33.4 kJ/mol to 166.4 kJ/mol, and the pre-exponential factor from 0.1 to 6.8x109 1/s.
Other examples of the comprehensive models are Atreya’ s one-step global model (Atreya
1984), Parker’ s one-stage multi-reaction model (Parker 1985), DiBlasi’ s two-stage semi-
global reaction model (DiBlasi 1993).
3.1.2 Typical pyrolysis models
As mentioned above, Quintiere’ s integral thermal model was applicable for the
one-dimensional unsteady conduction problem (Quintiere and Rhodes 1994; Spearpoint and
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
67
Quintiere 2001). It has all of the major features described previously. For example, the
surface pyrolysis was triggered at a fixed pyrolysis temperature. It is suitable for predicting
pyrolysis process of thermoplastics, such as PMMA. With some modifications, it has been
applied to model the wood pyrolysis in a cone calorimeter environment (Spearpoint and
Quintiere 2001). A degree of agreement to the experimental data has been achieved.
However the application of this model to a situation consisting of complicated physical and
chemical reactions is still questionable since the simulated details in pyrolysis are limited.
Janssens developed a comprehensive pyrolysis model for the investigation of wood structure
performance exposed to fire (Janssens 1994; 2004). The model is based on one-dimensional
heat transfer analysis in the wall and assumes a zero value for the heat of pyrolysis.
Moisture content was taken into consideration but not the moisture movement to the cold
side. The significant feature of this model is that it takes char contraction into account. This
function was achieved by adopting a sub-model named CROW (Charring Rate Of Wood).
The CROW model is based on White’ s correlation (White and Nordheim 1992) for charring
rate under standard fire conditions, according to the ASTM E119 test. Generally this model
provides an ability to predict the charring rate of wood exposed to the specified fire
conditions. Additional comparison and validation are needed to extend its application range.
Yuen et al. (1997) developed a three-dimensional mathematical model for studying
pyrolysis of wet wood. This model includes detailed considerations of evaporation of
moisture, anisotropic and variable material properties, and pressure-driven internal
convection of gases in wood. The 3-D modelling capability enables it suitable for prediction
the pyrolysis of a wooden cube inside a furnace under various temperatures. The pyrolysis
reaction is modelled by a first-order Arrhenius reaction. Totally six first-order reactions
representing the competing thermal decomposition reactions of various constituents have
been formulated in the computational code for future applications.
The kinetics for the pyrolysis of the single constituent (in a beech wood) was obtained from
a research conducted by Bonnefoy et al. (1993). The kinetics was determined by a method
of best fit of mass loss to the experimental results for the wooded cube heated in a furnace.
The obtained values for the kinetics are E = 125 kJ/mol and A = 1.25×108 1/s. Yuen et al.
found that the computed mass loss is lower than the experimental data within the first 35
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
68
seconds in a 973 K environment. According to the researchers, the use of the single kinetic
reaction for the modelling is the reason. It is known that competing primary reactions due to
the different major constituents (i.e., cellulose, hemicellulose and lignin) and secondary
reactions exist in wood pyrolysis (Bradbury et al. 1979). They believed that the prediction
might be improved by modelling the multiple competing reactions. The method to
determine the kinetics for the total 6 reactions corresponding to the major constituents was
developed by Alves and Figueiredo (1989) which conducting series isothermal TGA tests.
However, no applicable results have been published for adopting such method due to the
complication of simulating those reactions properly under various heating conditions.
Atreya’ s one-dimensional heat transfer pyrolysis model is a comprehensive one-step global
reaction model. It is based on Atreya’ s earlier research on wood pyrolysis and horizontal
flame spread (Atreya 1984). This model solves the heat conduction equation with locally
varying density, thermal properties and reaction rates to compensate for char formation.
With continuous improvement by the author and other researchers, it has the built-in ability
to model most aspects of pyrolysis phenomena, such as charring, moisture content, etc.
Some developments of this model have been performed and several applications of this
basic pyrolysis model have been published, which will be further reviewed later in Section
3.3.
3.1.3 Criteria of model selection for current research
As indicated in the review performed by Moghtaderi (2001), the major weakness of the
current thermal model is the lack of detailed chemical pyrolysis expression and poor
accuracy in predicting details related to chemical pyrolysis. For example, the integral
thermal model could not accurately predict the onset of pyrolysis. This is because the
initiation of pyrolysis that occurs at a low temperature range is controlled by chemical
kinetics rather than thermal transport. However, this does not mean that the most
complicated comprehensive models, such as two-stage multi-reaction and two-stage semi-
global, are automatically the best. It is because that there are a great number of details about
the semi-reactions which are still unknown or difficult to describe. For example, in the
previous isothermal TGA tests developed by Alves and Figueiredo (1989), the temperatures
to separate various reactions are difficult to determine. To handle a huge number of details
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
69
in today’ s pyrolysis models is either impossible or unnecessary. Modelling is simplification.
As identified in the literature review, a solution for engineering applications is to include
enough major phenomena and appropriate details about the phenomena but with a robust
kinetic mechanism.
Meanwhile, when heat transport details, such as surface radiative heat loss, are taken into
consideration, the difference between experimental data and integral model’ s prediction will
be increased dramatically (Moghtaderi 2001). These shortcomings led to the elimination of
the thermal models from consideration as a modelling tool for pyrolysis and ignition.
To meet the research requirements specified in Chapter 1, a set of criteria for choosing the
pyrolysis and combustion model was identified by the author. Except for normal
requirements like high accuracy and sufficiently fast computational speed, the following
criteria are of major concern:
including major pyrolysis details, such as charring, moisture content, etc
presenting multiple ignition criteria such as critical surface temperature and
critical mass loss flux
the ability to obtain distribution of gas phase products
the ability for users to intervene, especially with the kinetics, is highly
desirable, as this will provide the possibility to model more closely the
materials studied in this thesis.
Among these criteria, user control of kinetics is essential for the current research and the
description of gas phase fuel distribution is a preference.
Against the selection criteria above, Atreya’ s model comes out as the most satisfactory
candidate. Atreya’ s model includes the key functions in the study of pyrolysis. The one-step
global pyrolysis model makes it a suitable engineering tool. Various improvements,
especially a robust numerical computational scheme in a computer model (McGrattan et al.
1998), have been achieved recently. This computer model has been widely applied in FSE.
When balancing the advantage of robust computational ability and the disadvantage of a
lack of minor phenomena, the author believes that Atreya’ s model is a suitable prototype for
modelling pyrolysis and ignition processes. Therefore this computer model based on
Atreya’ s model was chosen in this research. Further discussion of Atreya’ s model is
presented in the following section.
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
70
3.2 Atreya’s One-dimensional Heat Transfer and Pyrolysis Model
Based on the basic pyrolysis process described in the literature review, Atreya developed a
one-dimensional heat transfer and pyrolysis model (Atreya 1984). The major content of this
model is presented briefly in this section.
3.2.1 Basics of Atreya’ s pyrolysis model
3.2.1.1 Physical model
The physical configuration of the system is illustrated in Figure 3.1.
Figure 3.1 Physical configuration of the case being studied
Definitions for the variables used in Figure 3.1 are as follows:
F: total cold wall heat flux, kW/m2;
T: solid temperature, K;
Ts: surface temperature, K;
T∞: ambient temperature, K;
t: time, s;
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
71
h: convective heat loss coefficient, kW/m2⋅K;
e: base of natural logarithms;
A: pre-exponential factor, 1/s;
E: activation energy, kJ/mol⋅K;
R: universal gas constant, =8.314×10-3 kJ/mol⋅K;
λ: conductive coefficient, kW/m⋅K;
cp: specific heat of solid, J/kg⋅K;
ρ: solid density of fresh solid, kg/m3;
ρc: density of char, kg/m3;
ρ∞: density of solid at ambient environment, kg/m3;
σ: Stefan-Boltzmann radiation constant, W/m2⋅K;
y: coordinate normal to the solid surface.
This physical problem can be described as below. A solid fuel slab, wood in Atreya’ s
research, initially at equilibrium with the surrounding atmosphere, is subjected at t = 0 to a
constant incident radiative heat flux F. It is assumed that the gas phase is inert (no oxidative
effects) and that the influences of solid-phase cracking, shrinkage, surface regression, and
grain direction are negligible. The interaction between F and the gaseous products of
pyrolysis is also neglected. Thus, F is assumed to be the time-average radiative heat flux
reaching the surface. The wood is assumed to be nondiathermic and opaque to the incident
radiative flux.
3.2.1.2 Four stages of wood pyrolysis
There are four stages of the wood pyrolysis process, as shown in Figure 3.2.
The initial heating phase, herein called the “ inert heating stage” , is characterized by the heat
of the surface to Tpy, the pyrolysis temperature, which is defined as the value of T at the
inception of pyrolysis. No chemical reactions occur in this stage, and since the material is
considered dry (i.e., there is only bound or hygroscopic water, see Section 2.3.3.2), there are
no effects of moisture. The change in density is negligible until the solid temperature is
close to the pyrolysis temperature.
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
72
Figure 3.2 Stages of wood pyrolysis
In the next heating phase (the initial pyrolysis stage or the transition regime) the initial
release of volatiles from the surface coincides with the formation of a pyrolysis front. In this
phase a rapid rise in the volatile mass flux is observed, and heat losses from the surface by
convection and re-radiation become important. The surface temperature and the volatile
mass efflux continue to increase until a thin char layer forms.
In the char layer, the temperature is defined as charring temperature, Tc. In this heating stage
(thin char) the mass flux reaches its maximum value.
The final heating stage (thick char) is characterized by a gradual decline in mass flux, m ′′ ,
and a correspondingly gradual increase in the char layer thickness. The expected time span
of each of these four stages is illustrated qualitatively in the m ′′ versus t plot of Figure 3.3.
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
73
Figure 3.3 Pyrolysis regimes in a qualitative plot of the volatile mass flux versus time
(source: Atreya (1984))
3.2.2 Basic equations and assumptions
Three important assumptions were made in Atreya’ s study. The first assumption is that the
thermal interactions between the volatile gases and the reacting material can be neglected. It
is known that the generation of volatile gases inside the solid sample can produce high
pressures (up to 0.3 atm (Lee et al. 1976), depending on the wood porosity). And they force
the volatiles toward both directions, to the hot char layer and to the interior of the solid
condense, only to be subsequently re-gasified. The net heat transfer between the volatiles
and the hot char, whose magnitude is measured by the quantity cp,g(Ts-Tc), (cp,g is the
specific heat of the volatiles), is ignored for two reasons. First, for small char layer thickness,
cp,gTc(Ts/Tc-1) is negligible because (Ts/Tc-1) is far less than 1. Second, when the char layer
thickness is large, and Ts/Tc far larger than 1, most of the volatiles are issued through cracks,
which were observed in the char layer before ignition in most experiments (Goos 1952;
Tang 1972). This makes the net heat transfer between volatiles and char negligible. The
effects of condensation and re-gasification are also ignored since the temperature gradient is
decreased down from the surface and these effects will be very weak while volatiles are
releasing.
The second assumption is that the chemical processes occurring in the decomposition of
“ dry” wood to char can be modelled by a single, one-step rate equation containing three
fixed parameters, the char density ρc, the pre-exponential factor A, and the activation energy
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
74
E. This assumption is an idealization of the actual process since wood decomposes in a
complex manner, producing hundreds of compounds. However, there is presently much
controversy concerning its precise nature and no generally accepted reaction pathway,
analogous to those used to model gas phase reactions, has yet been developed.
The third assumption is that the heat of pyrolysis, pyH∆ , in Atreya’ s original model is
negligible. As discussed by Atreya (1984), there is much confusion about this term.
Reported values vary between 750 J/g (endothermic) and -18800 J/g (exothermic). The
exothermicity apparently arises from secondary reactions between the oxygen and the hot
char, or the pyrolysis gases. However, the net exothermicity of the process must be
relatively small, since thermal runaway (i.e., explosion) has never occurred during wood
pyrolysis. An estimate for the upper bound of pyH∆ is cp(Tpy-T∞) ≈ 0 (418 J/g), which is the
sensible heat required to raise 1 g of wood to its pyrolysis temperature. Thus, pyH∆ < 418
J/g could lead to thermal runaway. Therefore, Atreya argues that whether exothermic or
endothermic, pyH∆ is generally much smaller than the thermal diffusion term in the
equation for the conservation of energy (see Figure 3.1), and pyH∆ ≈ 0 is assumed here.
This assumption has been employed by previous researchers (Atreya 1984; Parker 1985).
With the above assumptions, the model equations and boundary conditions describing the
pyrolysis of wood can be reduced to those shown in Figure 3.1. The density and thermal
conductivity of timber are all functions of solid temperature. If it is also assumed that the
thermal conductivity is proportional to the density, i.e., λ= ρk, where k = constant = λ∞/ρ∞,
then the equations from this figure can be re-written as follows. It is noted that all the
variables in following equations have same definitions as in Figure 3.1.
Heat transfer equation at surface:
)()( 44
∞∞ −−−−=∂∂− TTTThF
yT σλ
(3.1)
Within the solid:
Energy:
∂∂
∂∂=
∂∂
yT
ytT
cp λρ (3.2)
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
75
Decomposition: ( ) RTEc eA
t/−−−=
∂∂ ρρρ
(3.3)
Mass:
tym
∂∂=
∂∂ ρ
(3.4)
Boundary and initial conditions:
( ))1(),0(
0,
),()0,(
−Γ−=∂∂−
==∞=
∞
∞
∞
sTTtyT
y
TtTyT
ρ
ρρ
(3.5)
where:
Ts: surface temperature, K;
m: mass of the solid, kg;
Γ = hT∞/F + (σT∞4/F)(Ts + 1)(Ts
2 + 1), the linearized convective and radiative heat
losses term.
The derivation for the critical parameters, such as pyrolysis temperature, pyrolysis time,
charring temperature, charring time, and mass flux at different stages, are presented in
Appendix B.
3.3 Improvements and Applications for Atreya’s Model
Major improvements to Atreya’ s model were made by Parker (1989), Steckler et al. (1994),
and Ritchie et al. (1997). The improvements include incorporation of charring properties,
backing substrate, and moisture content, etc. The works of Parker and Ritchie et al. are
described in the following two sub-sections.
3.3.1 Improvement by Parker
In his numerical scheme, Parker (1985) divided the fuel slab into thin isothermal slices, and
calculated mass flux on each slice. The total mass loss was summed from all the slices. The
shrinkage was calculated from the char contraction coefficients. Moisture content was taken
into account in the energy equation. The ignition criterion was the critical mass flux, set as
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
76
2.5 g/m2⋅s, excluding the release of moisture. When the calculation results were compared
with cone calorimeter testing data, a reasonable agreement was achieved (Parker 1992).
3.3.2 Improvement by Ritchie et al.
Ritchie’ s modified model (Ritchie et al. 1997) includes descriptions for conduction of heat
to the back substrate materials, the evaporation of moisture and the decomposition of the
virgin wood into gas phase fuel products and char. In other words, Atreya’ s third
assumption was eliminated. The time lag for the movement of gases products within the
solid, which is caused by the obstacles, is ignored. This assumption results in an
instantaneous release of the gaseous products.
On the surface, the governing equation for energy (Equation (3.2)) is detailed as below:
( )[ ] ( )[ ]∞∞ −−∆∂
∂+−−∆∂
∂+
∂∂
∂∂=
∂∂
TTCHt
TTCHty
Tyt
Tc ev
mpy
ww 21
ρρλρ (3.6)
where:
wρ : the total density of the wood, kg/m3;
mρ : the moisture density, kg/m3;
pyH∆ and evH∆ : the heat of pyrolysis of wood and the heat of evaporation of water,
kJ/kg;
wλ : conductivity for the wood, W/m⋅K, and determined by Equation (3.7);
C1 and C2: coefficients and given by Equations (3.8) and (3.9);
cρ : average density and specific heat for pyrolysis process, kg/m3, and
T∞: ambient temperature, K.
The boundary condition on the wood surface is due to radiation and convection, and reaches
a balance as follows:
gasgrad
solidw y
Tq
yT
∂∂−=
∂∂− λλ '' (3.7)
where: ''
radq : the net radiative heat flux on the surface, kW/m2, and
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
77
gλ : conductivity for the gas phase, W/m2⋅K.
The two coefficients C1 and C2 are defined as
gp
wfw
wfpwfwpw ccc
C ,0
,0,01 −
−−
=ρρρρ
(3.8)
gpmp ccC ,,2 −= (3.9)
where:
0wρ and wfρ : densities of the virgin wood and the char respectively, kg/m3;
0,wpc and wfpc , : average specific heats of the virgin wood and the char during the
pyrolysis period respectively, kJ/kg⋅K;
gpc , and mpc , : average specific heats of the gaseous products and the moisture
during the pyrolysis period respectively, kJ/kg⋅K.
Inside the solid timber, the pyrolysis rate is modelled as a first order Arrhenius reaction:
RTEa
w Aet
/−−=∂
∂ ρρ (3.10)
where:
aρ : density of the active wood, kg/m3;
A: the pre-exponential factor, 1/s, and
E: the activation energy, kJ/mol⋅K.
As a simplification, equation (3.10) is only applied after the wood temperature reaches
100 oC. The evaporation of moisture is assumed to consume all the available energy while
the wood reaches the evaporation temperature, Tev. The evaporation rate is given by
following equation:
( )[ ]
evTT
wm
ev yT
ytTTCH
=∞
∂∂
∂∂−=
∂∂−−∆ λρ
2 (3.11)
The thermal properties of the wood during the moisture releasing and charring processes are
calculated by the following equations:
( )mpmwfpcwpa cccc ,,0, ρρρρ ++=
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
78
+
=
wf
cwf
w
aww ρ
ρλρρλλ
00 (3.12)
wfw
wfwwa ρρ
ρρρρ
−−
=0
0
awc ρρρ −=
where the thermal properties of the virgin wood and the char, 0,wpc , wfpc , , 0wλ and wfλ are
functions of local temperature and have been described by other researchers (Hostikka
2002).
Combined with the gas phase sub-model, Ritchie’ s global analytical model was used to
predict the Douglas fir burning in bench-scale tests (Ritchie et al. 1997) and reasonable
agreement has been achieved.
3.4 Calculation for a Studied Material Based on Atreya’s Model
3.4.1 Thermal properties of studied material
The selection of appropriate material properties and pyrolysis kinetics is a very difficult task.
Some of the properties were measured from the experiments described in previous chapters,
but most of the values were sourced from published results. Where no data for the identical
testing materials are available, parameters extracted from similar materials are included. The
pre-exponential factors and activation energies were verified with the experimental data
obtained in the current study. The thermal properties for one of the tested woods, mountain
ash, are listed in Table 3.1.
Table 3.1 Thermal parameters for the studied timber, mountain ash
Parameter Value Unit Source ρc 146.5 kg/m3 Measured by the author ρw 681.3 kg/m3 Measured by the author λwo 3.054x10-7T∞+3.62x10-5 kW/m⋅K (Hostikka 2002)
cp,∞w 0.01+0.0037T∞ kJ/kg⋅K (as above) α∞ 1.94x10-7 m2/s (as above)
∆Hev 125.6 kJ/kg (as above)
sT (T∞+Tig)/2 K Tig =350 oC was chosen for normal timber materials
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
79
The inputs for Atreya’ s model are given in Table 3.2. The evaluation of these input
parameters is included in Appendix C.
Table 3.2 Inputs for calculation by Atreya’s model L 7.38x10-4 m τ 2.81 s F 50 kW/m2 A 1.21x108 1/s E 140.2 kJ/mol Γ 0.4
3.4.2 Calculation results from Atreya’ s model
Manual calculation was performed following a procedure provided by Atreya, which is
listed in Appendix C, and results are presented therein. A plotting for the non-dimensional
mass flux curves is also shown in Figure 3.4. The variables shown in this figure are non-
dimensional mass flux, m*, and non-dimensional time, t*.
0
0.5
1
1.5
2
0 1 2 3 4
t*
m*
Thin char
Thick char
Figure 3.4 Non-dimensional mass flux for mountain ash under 50 kW/m2
Compared with Atreya’ s calculation result for inputs of a construct timber fuel, (shown in
Appendix C), current calculation results for mountain ash has a similar trend in mass flux
plotting. It is shown that in the thin char pyrolysis stage, the mass flux increased by an
exponential curve following the Arrhenius equation. As the pyrolysis process changed into
Chapter 3 Modelling of Thermal Decomposition and Ignition of Solid Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
80
the thick char stage, the increased thickness of the charring layer starts impeding the heat
and mass transferring through the charring layer. Therefore the mass flux curve decreased
from its peak value. This mass flux plotting can be used later to compare with the Fire
Dynamics Simulator (FDS) model’ s simulation results.
3.5 Summary
A combustion model is a critical component in a fire model. A practicable combustion
model should be able to model pyrolysis and ignition phenomena properly. All available
pyrolysis models are simplified descriptions for the pyrolysis process. Some are quite
simple while some are relatively complex and able to deal with detailed phenomena and
reactions. An engineering approach can be achieved by balancing the details a model can
handle and the computational ability available. Combined with the research aims specified
in Chapter 1, a set of criteria for selecting pyrolysis models for this research has been
identified.
Major types of existing pyrolysis models have been discussed in this chapter. Atreya’ s one-
dimensional heat transfer and pyrolysis model was chosen as modelling tool according to
the criteria developed. Atreya’ s model, including its physical problem, assumptions, basic
equations and theoretical calculation, has been discussed in detail. Major improvements for
the model by a number of researchers have also been presented. These improvements
involve a numerical scheme, shrinkage of wood during combustion, moisture content, etc.
Following Atreya’ s model, theoretical calculation was also performed for mountain ash, one
of the furnishing materials under study. The calculation mass flux matches the trend from
Atreya’ s original results.
81
STUVWXY!ZSTUVWXY!ZSTUVWXY!ZSTUVWXY!Z X[OV\X&Y]B^DX_ W.U`'a=XSbM^IVJbX[OV\X&Y]B^DX_ W.U`'a=XSbM^IVJbX[OV\X&Y]B^DX_ W.U`'a=XSbM^IVJbX[OV\X&Y]B^DX_ W.U`'a=XSbM^IVJbOc]%W]LbM_c]%W]LbM_c]%W]LbM_c]%W]LbM_c)W)daeGfbY WTXgfhd*Y&_J]icc)W)daeGfbY WTXgfhd*Y&_J]icc)W)daeGfbY WTXgfhd*Y&_J]icc)W)daeGfbY WTXgfhd*Y&_J]icjTk]_IlTk]_IlTk]_IlTk]_Il^U'WX&Y]PUQ`Rc^U'WX&Y]PUQ`Rc^U'WX&Y]PUQ`Rc^U'WX&Y]PUQ`Rc
This chapter involves the application of the basic thermal experiment, the TGA test. The test
method and computational scheme for obtaining decomposition kinetics are examined and
improved according to the basic analysis for decomposition that was performed in Chapter 2.
Then this modified scheme is used to obtain the “ effective” kinetics for the studied materials.
Three sections are included in this chapter. The first section introduces the test device, its
principle and test method. The computational scheme for the kinetics is also analysed and
improved. In Section 4.2, the test procedure and test results for the materials included in the
current study are presented and discussed. The last section, Section 4.3, summarises the test
results and their potential application.
4.1 TGA Experiment and its Application
As reviewed in Chapter 2, the TG analyser is a commonly used tool for basic
thermodynamic studies due to its wider temperature testing range and environmental gas
control capability. It is easy to analyse the thermal pyrolysis processes of solid materials
under confined environments. This experimental device and computational schemes have
been widely adopted by researchers to determine kinetics for various materials.
4.1.1 TGA apparatus, principle and test procedure
The TGA analyser was designed to investigate the thermal characteristics of a tiny amount
of samples. It consists of:
a radiant heating chamber (furnace)
a quartz reactor
a temperature controller
a precision balance
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
82
a gas feeding system
an acquisition data set
An illustration of the major structure of the TG analyser is shown in Figure 4.1. The sample
to be tested is exposed to thermal radiation in a platinum basket and the temperature is
continuously recorded by a chromel-alumel thermocouple (0.1 mm bead) placed under the
basket with a 0.1 mm gap. A precision balance (accuracy 0.1 to 1 g) allows the sample
weight to be measured while a continuous air flow establishes the proper reaction
environment. The gas flow also carries away the volatilisation products, thus minimizing
any secondary reaction activity. The testing procedures involve either constant temperature
ramp or constant temperature experiments. The sample size and heating condition can be
chosen so as to avoid significant deviation with respect to the desired temperature value.
The process can also be carried out with a continuous control of the sample temperature
through variation of the intensity of the applied radiative heat flux.
Figure 4.1 Illustration of a TG analyser
One standard usage of the TG analyser is the so-called “ dynamic” or “ non-isothermal”
experiment. In this kind of experiment, a sample is heated at a desired heating speed,
normally a constant heating rate or a linear time-temperature curve. The heating rate can
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
83
vary from 0.05 K/min to 200 K/min. Much higher heating rates, up to 100 K/s, may be
achieved by using special techniques.
Alternatively, the isothermal experiment can be conducted. Samples are first heated to the
required temperature very quickly, at a rate higher than 700 K/min. Then the sample mass
and mass loss rate can be monitored at the required temperature for a specified time
duration.
Both dynamic and isothermal methods can be used in oxidative or inert gas flow
environments. The test results from these methods can be used for calculation of the
kinetics. The computational methods are discussed in the following sub-section.
4.1.2 Research methodology
From the TGA test, two kinds of plots are available. One is the plot of sample weight (or
weight fraction) against temperature: the thermogravimetric (TG) curve. The other is a plot
of the rate of sample weight loss against temperature: the differential thermogravimetric
(DTG) curve. Obtained under a set of standard conditions, the TG and DTG curves provide
information of the combustion reactivities of the fuel from the onset of oxidation to
complete burn-out (Wagoner and Duzy 1976). Meanwhile other parameters, such as the
kinetics of a reaction, can be derived from these curves (Seungdo and Park 1995).
4.1.2.1 Curve trend and characteristic temperatures
The TG and DTG curves were at first used to present burning profiles of fuels. The
“ burning” process is generally characterized by several events (Wagoner and Duzy 1976).
For example, in an oxidizer flow environment TG test for coal, as well as for many other
solid fuels, the following events may occur. The initial mass loss rate peak in a DTG curve
at temperatures below 100 °C corresponds essentially to the release of moisture. The rate
then falls at about 200-300 °C. This is caused by oxygen absorption by the solid before
subsequent combustion. The main combustion peak is usually preceded by a small peak or
shoulder which is partly attributable to volatile release. Then the mass loss rate reaches a
peak value (MLRmax) and the corresponding temperature is the peak temperature (Tmax).
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
84
An example of TG/DTG plotting for a timber material can be found in Figure 4.2.
0
20
40
60
80
100
0 100 200 300 400 500 600
Temperature (oC)
Wei
gh
t (%
)
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
Mas
s L
oss
Rat
e (m
g/m
in)
Weight (TG)
MLR (DTG)
Figure 4.2 A typical TG (WT curve) and DTG (MLR curve) plotting
The peak temperature Tmax can be used to assess the relative combustibility. A higher Tmax is
indicative of a less reactive fuel, and vice versa.
4.1.2.2 Kinetics and calculation method
In many kinetic formulations of solid state reactions, it has been assumed that the isothermal
homogeneous gas or liquid phase kinetic equation can be applied (Baker 1978). An
expression of Arrhenius equation for a one-order kinetic reaction has been shown in Section
2.3.2. Generally, for a single kinetic reaction, a DTG curve (as expressed as conversion rate
H (dx/dT)) can be represented by the following Arrhenius equation based on reaction order
(Seungdo and Park 1995):
nRTE
xeA
dTdx
)1( −=
−
β (4.1)
where:
x: weight fraction of conversion, and
∞−
−=
WWWW
xI
I
W: sample weight, g;
WI: initial sample weight, g;
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
85
W∞: end sample weight, g;
T: sample temperature, K;
A: pre-exponential factor, 1/s;
: heating rate, K/s;
E: activation energy, kJ/mol;
R: universal gas constant, =8.314×10-3 kJ/mol⋅K;
n: reaction order.
At the peak temperature Tmax, differentiating the right hand side of Equation (4.1) with
respect to temperature yields zero:
0)1(
max
1max
2max
=−−
−
−RT
E
n eA
xnRTE
β
(4.2)
where:
Tmax: temperature at peak conversion rate, K;
xmax: sample weight fraction of conversion at peak conversion rate.
A great number of methods have been developed to compute kinetics from the TGA tests.
The first example was illustrated here. The peak temperature (Tmax) and corresponding
maximum sample weight fraction of conversion (xmax) and the maximum conversion rate
(Hmax) can be utilized to estimate the activation energy and pre-exponential factor of a
reaction. Substituting Equation (4.1) into Equation (4.2) yields the expression for the
activation energy and the pre-exponential factor
max
max2
max
1 xHnRT
E−
= (4.3)
and then replacing Equation (4.3) into Equation (4.2)
n
RTE
xeH
A)1( max
maxmax
−=
β
(4.4)
where Hmax is the peak conversion rate, max
max
=dTdx
H .
If the reaction order, n, is known, then the activation energy, E, and the pre-exponential
factor, A, can be estimated from experimental measurements of Tmax, Hmax and xmax.
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
86
To obtain those three parameters, Coats-Redfern method has been reported as the most
frequently used (Vyazovkin and Wight 1998). Equation from the Coats-Redfern method was
given as follow:
RTE
EAR
Txf −
=β
ln))(
ln( 2 (4.5)
where f(x) is a function that represents the reaction model. When Arrhenius equation was
assumed, expression from Equation (4.1) can be applied, f(x) = (1-x)n.
Therefore Equation (4.5) can be rewritten as
( )
( ))1(,
1ln
)1(,1
111
2
2
1
==−−
≠=
−−−
−
−−
neE
ART
x
neE
ARTn
x
RTE
RTEn
β
β
(4.6)
A plot of ln[-ln(1-x)/T2] vs 1/T (for n=1) or ln[1-(1-x)1-n]/[(1-n)T2] vs 1/T (for n1) gives
a straight line of slope –E/R. The pro-exponential factor A can be calculated from the
intercept of this straight line. The reaction order, n, was then determined from a series of
values that can give the best fit to the plots, i.e. the one generates smallest relative standard
deviation.
Another widely adopted method is the so-called Ozawa-Flynn-Wall (OFW) method (Ozawa
1970) which was based on traditional multiple heating rate technique. It has the advantage
without supposition of the reaction order n.
By linearizing Equation (4.1) the following form is obtained:
[ ]RTE
xAdTdx n −−= )1(ln)ln(β
(4.7)
This equation can be further reduced with Doyle’s approximation (Doyle 1962):
( )RT
ERAExf
05.133.5ln/ln)(ln −−−= β
(4.8)
where f(x) is a function of reaction model.
At each conversion level (x), the item ( ) ( )[ ]RERAExf /05.133.5/lnln −+− is constant. A
plot of logarithm heating rate (lnβ) vs reciprocal temperature (1/T) will yield a slope of
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
87
1.05E/R. By running multiple various heating rate tests, the values of activation energy, E,
can be computed at different sample fractions of conversion. This is the so-called OFW
method. An illustration of the application of the OFW method is given in Figure 4.3. Each
line in the figure was generated by a conversion level x under various heating rates.
Figure 4.3 OFW plotting of heating rate, ln(), vs temperature, 1/T,
for various conversion fraction, x
Furthermore, a special format of OFW method can be developed. Based on an observation
by Flynn and Wall (1966), for each peak in a DTG curve the xmax would rarely vary with
heating rate. Therefore the activation energy computed by a single mass conversion fraction
xmax was used for the reaction. This method could be considered as a special case of the
OFW equation. From Equation (4.4), following expression exists for different heating rate.
( )maxmax
max1lnln
RTE
HxA n
−
−=β (4.9)
Applying Equation (4.9) to two different heating rates, the activation energy can be obtained
by Tmax, Hmax and β from the two tests:
−⋅=
1max
2max
2
1
2max1max
2max1max lnHH
TTTT
REββ
(4.10)
Comparing above three methods, the method expressed in Equation (4.3) and (4.4) has an
advantage of requiring only single running of test at a heating rate. The disadvantage is that
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
88
the reaction order has to be estimated first. The estimation of the reaction order may affect
the accuracy of the obtained kinetics. Coats-Redfern method can calculate all the parameters
from a single test. However relatively complex mathematic computation is required. The
calculated results rely on the step of reaction order in the trial. The OFW method has the
advantage of not relying on the supposition of the reaction order n. By comparing nearly 20
typical computing models for kinetics, Carrasco (1993) found that OFW method has
satisfactory high accuracy. The special format OFW method has similar accuracy while the
peak conversion rate and peak temperature are required. Furthermore, the OFW computing
scheme can provide additional benefit. The kinetics obtained from a wide heating rate range
can be verified by the linearity of the plotting. For example, for a peak conversion fraction
(xmax), if two or more slopes (presenting the activation energy) were observed from the lnβ
vs 1/Tmax curve, different decomposition behaviours may exist under different heating
conditions. In the current study, the Coats-Redfern method was used to calculate kinetics
under very low heating rate, from 5 to 10 K/min, and therefore compared with results from
others. The OFW method was used to check decomposition behaviour and compute the
kinetics under relatively high heating rates for the studied materials.
4.1.2.3 Concern for test conditions and computational schemes
When the TGA test is used to study chemical and physical processes of solid samples, one
of the main drawbacks concerning solid phase chemistry is that a large part of the analyses
has been carried out in the presence of serious heat and mass transfer limitations. For inert
atmospheres, these mainly arise from the large sample size, so that significant temperature
gradients are established along the sample and the activity of secondary reactions is not
negligible. In the presence of oxidants, the process is further complicated by the strong
exothermicity of the combustion reactions. So that, apart from the wide variability of the
thermal conditions, the response time of the measuring devices may become too long
compared with the characteristic times of process evolution. Therefore some methods and
configuration are necessary in order to minimize these effects.
The configuration of the sample, the thickness and size, plays an important role, along with
the heating conditions. DiBlasi et al. (1999) found for a powdered wood sample that
ignition is a function of number of factors, such as sample size, heating rate, final heating
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
89
temperature, as well as gas flow condition. The sample size they suggested to avoid ignition
is between 2 to 25 mg. Either the heating rate or the final heating temperature has to be
controlled depending on the type of wood species.
Parker improved the TGA test method by introducing an extra oxygen analyzer to calculate
kinetics (Parker 1988). The test device and method have been briefly introduced in Section
2.6.1. This method is believed to eliminate the temperature gradient on the sample in a TGA
environment.
Some other efforts focused on improving kinetic computational schemes by TGA data.
Schultz et al. (1989) improved the calculation method by carrying out isothermal tests.
Kinetics were calculated through the Arrhenius equation by using measured reaction rate
and pyrolysis temperature. This isothermal test method is believed to eliminate temperature
lag complications.
The measured temperature inside the TG analyser was expected to differ from the real
temperature of the sample specimen due to the placement of the thermocouple. The
measured temperature is approximately in parity to the real temperature and the difference
may be proportional to the heating rate applied. The error in temperature measurement will
be translated into errors in the estimated parameters such as onset and maximum
temperatures. However, it is expected that the rate of temperature rise of the specimen
would closely follow the rate of measured temperature rise, i.e. a similar temperature
difference between the measured and real sample temperatures exists over the whole heating
process. ASTM also gives a normal temperature measuring error range of 20 oC
(ASTM_E1582 2000). As such, the quantities that are determined by a linear relationship
(the slope) of the heating rate and the measure reaction rate would be less sensitive to errors
in the temperature measurement. Seungdo and Park (1995) found that kinetic values are
more sensitive to slopes in the plotting of peak temperatures and peak reaction rates, not the
peak values themselves. This indicates that the kinetics may be less affected by the error in
temperature measurement.
As indicated in Section 2.3.2, the activation energy, one of the major parameters of kinetics,
is a material property and its intrinsic value believed to be independent on reaction
conditions, for example the heating condition. To obtain this intrinsic value above efforts
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
90
have being made to improve test method. However, due to limitations exist in every test
method, only “ perfect” values, which being considered very close to the intrinsic value but
approximate expression, can be achieved. To obtain this “ perfect” kinetics, small size and
tiny amount of specimen should be tested under very low heating rate.
On the other hand, the “ perfect” kinetics was computed from test data according to various
reaction models. For the same decomposition process, kinetics obtained by different
reaction models may vary. As studied by Vyazovkin and Wight (1998), major issues in
computing schemes are as follows:
Arrhenius equations may be meaningfully applicable only to reactions that take
place in a homogeneous environment
For non-isothermal TGA tests, solid decomposition processes ordinarily show multi-
step kinetics that readily change with heating condition.
It’ s further argued that there are serious limitations to apply those “ perfect” kinetics directly
into description of decomposition of solid materials in real fire conditions. In real fire
environments, the “ drawbacks” existed in TGA tests are always present, such as
temperature gradient, heat and mass transfer within the solid. The so-called “ perfect”
kinetics are either difficult to obtain or unhelpful in describing the decomposition in a real
fire. In a later part of this chapter, the author will provide more evidence from analysis of
computed activation energies to support this argument.
Therefore, the author aimed to obtain some kind of simplified kinetics, “ effective” value,
which is suitable to describe decomposition process in a simplified modelling scheme, as
discussed in Section 3.1. This engineering approach for pyrolysis research is for solving real
fire problems with enough physical and chemical details by the applicable kinetics. In late
sections of this chapter, kinetics obtained from normal and the improved schemes will be
presented for further pyrolysis modelling.
4.1.2.4 Relationships among the parameters
There are two commonly used approximations for the parameters. From Doyle’ s (1962),
one expression is provided as below:
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
91
−−
+−+
=≠
−−−
==
maxmax
maxmax
05.133.51lnlnln;1
05.133.5lnlnln;1
RTE
nH
nR
AEn
RTE
HR
AEn
β
β
(4.11)
Kissinger (1957) assumed the followings:
( )nnxn
exn−−=≠
−==1/1
max
max
1;1
/11;1
(4.12)
By introducing those approximation schemes into Equations (4.3) and (4.4), Seungdo and
Park (1995) gave the following equations:
;00516.1
2max >
=∂
∂ξββE
RT
(4.13)
;08085.1max <
−=
∂∂
EeRH ξ
ββ (4.14)
where
.3305.5ln −
=
βξ
RAE
2max
)1/(
max
2max
max
;1
;;1
RTEn
Hn
eRTE
Hn
nn −
≈≠
≈=
(4.15)
From Equation (4.15), the following equation can be obtained: 5.0
maxmax
1
≈ HEeR
T (4.16)
Substituting Equation (4.16) into Equation (4.9) yields:
5.0max
max
max )1(lnln H
ReE
HxA n
⋅−
−≈β (4.17)
From Equation (4.13) and (4.14), it can be seen that the peak temperature increases with the
increased heating rate, but the peak conversion rate follows the reverse trend. Meanwhile,
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
92
linear relationships are expected between the peak conversion rate and inverse square of the
peak temperature, between lnβ and 5.0maxH (from Equation (4.16) and (4.17)). These
theoretical relationships are helpful to validate and group the experimental data.
4.2 Current TGA Test Results
In this section, specifications of the tested materials and experimental equipment are
described. The basic thermal decomposition processes from the TGA tests are presented.
Finally, the calculated thermal dynamic parameters for the tested materials are given.
4.2.1 Experimental materials and apparatus
4.2.1.1 Details of materials
The tested materials included charring and non-charring materials. They were polyurethane
(PU) foams, fabrics, and woods. Stamina, (a brand) PU foam is manufactured by Dunlop
and has a density between 29 to 35 kg/m3. Details of the tested materials are listed in Table
4.1. All densities listed were measured after 48 hours conditioning at 50% relative humidity
and 25 oC. The fabric samples were cut into a round shape with a diameter of 3.5 mm. The
woods were chipped into slices of 3×3 mm with various thickness from 0.2 to 0.5 mm to
obtain different sample weights. The PU foams were prepared as 3×3×2 mm dices. The
sample weights varied between 0.5 to 3 mg.
Table 4.1 Properties of the tested materials
Materials Description Density (kg/m3)
Hardwood mountain ash 681.3
Pinewood Australian pine 440.7
Standard PU foam A23-130 22.8
Stamina PU foam HR32-80 31.2
Cotton (fabric) 100% cotton 0.42 (kg/m2)
Cotton & polyester (fabric) 45% cotton and 55% polyester 0.37 (kg/m2)
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
93
4.2.1.2 Apparatus and test conditions
The experimental apparatus was a PerkinElmer TGA-7 analyser in the Applied Chemistry
Department, RMIT University, Australia. For the description of this apparatus, refer to
Section 4.1. In all tests the standardized flow rate was set at 200 ml/min.
The heating rates investigated were from 5 to 200 K/min in the dynamic experiments. A
constant temperature heating process, isothermal scanning, was also carried out. A gas
feeding system can provide varying gas flows, from inert to oxidizing atmospheres. In these
experiments, pure nitrogen and air flow were used.
4.2.1.3 Calibration
Calibrations for the TG analyser were performed for weight and temperature. While there
are no standards for those calibrations, it is a usual way to follow standard procedures
provided by the manufacturer. Since the calibration of weight is relatively simple, only
calibration of the temperature was described below.
Standard ferromagnetic samples with different Curie points were used in the temperature
calibration (Brown 1988). Those samples were placed in the balance and a magnet outside
the furnace alerted the total weight of the sample. When the temperature increased above the
characteristic Curie point of the specimen, they lost their ferromagnetic properties and were
no longer attracted by the magnet. This resulted in a sudden change in weight. The use of
number of ferromagnetic materials extends the calibration over a wide temperature range.
By checking recorded temperature against the characteristic temperatures of those
calibration specimens have, temperature lag can be then obtained. The supplied
ferromagnetic samples and their characteristic temperatures are Ni (355.5 oC) and Perkalloy
(596.0 oC). This calibration was carried out before changing heating rate. The temperature
lag was found less than 2 to 20 K for heating rates ranging from 10 to 100 K/min, which
met the requirement from the manufacturer (Perkin_Elemer 1991).
4.2.2 Measurement of thermal characteristics of the materials
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
94
Presented in this part are the decomposition processes of the testing materials as well as the
basic thermal characteristics such as mass loss rate, conversion rate and peak temperature.
4.2.2.1 Decomposition process
Figure 4.4 shows the TG and DTG curves for all six materials tested in the air flow at a
heating rate of 10 K/min. There is a moisture release phase under 160 oC for the woods and
fabrics, which shows in the DTG curves as the first small peak. Above this temperature,
there is one pyrolysis reaction for both woods and the cotton fabric, and two reactions for
the other materials.
In the temperature range from 500 to 600 oC, there is another “ reaction” for both woods.
This reaction is attributed to combustion of the charring residue, and is verified by
comparing the DTG curves in air and nitrogen flow conditions. Figure 4.5 (a) and (b) shows
the difference between these two gas flow conditions under a 20 K/min heating rate. It is
found that under the nitrogen flow condition, the second major peak in the DTG curve
disappears. The residue of sample weight (shown in the TG curves) under nitrogen flow is
about 10%, compared with almost zero for that under air flow. Meanwhile, for the
combustion of char under air flow the condition may also be affected by the heating rate,
which will be discussed in the following section. The DTG curve for cotton fabric also
shows a single peak. The approximate single peak pattern enables the one-step global
pyrolysis model, which was discussed in Chapter 3, to be applied to these materials.
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
95
Figure 4.4 TG and DTG curves on 10 K/min heating rate under air flow
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
96
Figure 4.5 TG and DTG curves for mountain ash under 20 K/min heating rate
The double reaction in the DTG curves in the cotton and polyester test (in Figure 4.4 (d)) is
a result of the presence of two components, cotton and polyester. Comparing the DTG
curves for the cotton (Figure 4.4 (c)) and the cotton and polyester fabrics (Figure 4.4 (d)), it
is found the first peak in DTG curve for the cotton and polyester has a similar temperature
range with the peak of the cotton. Thus the first peak in the cotton and polyester test is
attributable to the decomposition of the cotton component.
Some characteristic temperatures, Tmax, and maximum mass loss rates, MLRmax, for the
cotton fabric, mountain ash and the standard PU foam at various heating rates are listed in
Table 4.2. These results average values from 3 tests. A full list of all the TGA tests,
including other materials under study, can be found in Appendix A.
It can be seen from the Table 4.2 that for shown materials, under two types of gas-flows, the
peak mass loss rate and corresponding peak temperature increased as the heating rate
increased. Normally the tests in the air flow have lower peak temperature but higher mass
loss rate than that of in the nitrogen flow under same heating rate. The standard PU foam
shows two peak mass loss rates in most tests except the low heating rate in the nitrogen
flow.
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
97
Table 4.2 Peak temperatures and mass loss rates
Material Gas flow
Heating rate (K/min)
Tmax,1
(oC) MLRmax,1
(mg/min) Tmax,2
(oC) MLRmax,2
(mg/min)
10 367 1.24 -# - 50 440 6.36 - - Air
150 480 11.27 - - 10 394 0.75 - -
Cotton
N2 20 401 1.62 - - 10 350 0.48 - - 50 470 1.53 - - Air
200 516 6.36 - - 10 366 0.33 - -
Mountain ash
N2 30 369 1.24 - - 10 285 0.34 312 0.22 50 381 0.63 452 0.54 Air
200 417 1.27 470 1.74 20 309 0.22 - -
Standard PU foam
N2 30 313 0.31 358 0.42 # -: means that no second reaction was observed.
4.2.2.2 Effect of heating rate
The effect of heating rate can be seen from plotting of the conversion rate vs sample
temperature under various heating rates, in Figure 4.6 to Figure 4.9.
-0.030
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
0 100 200 300 400 500 600
Temperature (oC)
Con
vers
ion
Rat
e (1
/K)
10 K/min20 K/min30 K/min50 K/min100 K/min200 K/min
Figure 4.6 Conversion rates for the mountain ash under air flow and various heating rates
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
98
Figure 4.7 Conversion rates for the cotton under air flow and various heating rates
Figure 4.8 Conversion rates for the cotton and polyester
under air flow and various heating rates
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
99
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
0 100 200 300 400 500 600
Temperature (oC)
Con
vers
ion
Rat
e (1
/K)
5K/min10K/min20K/min30K/min
50K/min100K/min150K/min200K/min
Figure 4.9 Conversion rates for the standard PU foam
under air flow and various heating rates
For those materials that can be described by a single reaction, such as the woods and the
cotton, the peak conversion rate decreased and peak temperature increased with the
increasing heating rate. In fact, the above results were true for all peak temperatures and the
first peak conversion rates of all the materials tested under both nitrogen and air flow
conditions. This observation matches the previous theoretical analysis (Equation (4.13) and
(4.14)), as well as Seungdo and Jae’ s experimental results (1995). As indicated in Section
4.2.2.1, the second peak in mountain ash’ s DTG curve under air flow is caused by
combustion.
For the multiple reaction materials, the cotton and polyester and the standard PU foam, the
trend in the DTG curves under varying heating rates is quite different. For the standard PU
foam, the second peak conversion rate increases and becomes the dominant, under a high
heating rate range (from 50 to 200 K/min), when heating rate increased (Figure 4.9). The
conversion rate for the second peak for the cotton and polyester is quite stable in the low
heating rate range (from 10 to 50 K/min), but decreased in the high heating rate range (from
50 to 200 K/min) as the heating rate increased (Figure 4.8). This means that the effect of the
first reaction (i.e. the height of the first peak in the DTG curves) becomes gradually weaker
and the effect of the second reaction becomes stronger. This can be verified by activation
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
100
energy computation, which will be shown in Section 4.2.3.1, as well as Agrawal’ s results
(Agrawal 1992) which will be explained below. Some material has a higher activation
energy E for the first reaction followed by a lower E for the secondary reaction, such as the
standard PU foam shown in Table 4.3. For this type of material, Agrawal indicated that the
first peak conversion rate decreased and then is followed by a slowly increasing rate of
volatile evolution, which was limited by the low E reaction. At a high heating rate range, the
speeds of both the first peak height decreasing and the second peak height increasing
become slower as the heating rate increases. This shifting relation is helpful for predicting
the decomposition process under a high heating rate range, which is closer to real world fire
conditions.
Changes of the peak mass loss rates (MLRmax) followed the opposite trend of that of the
peak conversion rates. For example, a comparison between Figure 4.4 (b) and Figure 4.5 (a)
indicates that the mass loss rates of the mountain ash increased as the heating rate increased.
The peak mass loss rates of all testing materials increased with an increasing heating rate as
shown in Appendix A. However, the peak conversion rate decreased as the heating rate
increased, which can be observed from Figure 4.6 to Figure 4.9. The relationship between
parameters of conversion rate (dx/dT) and reaction rate (dx/dt) is derived from the following
equations, while heat rate β = (dT/dt):
dTdx
dTdx
dtdT
dtdx
dtdW
WWI
β=⋅==− ∞
1 (4.18)
or, at the peak temperature condition,
maxmaxmax
1H
dtdW
WWdtdx
I
β=
−=
∞
(4.19)
where WI and ∞W are sample weights at initial and end conditions of a test.
It has been seen that the peak mass loss rate value, max
dtdW
(i.e. MLRmax), is determined by
the product of β and Hmax. Even though Hmax (peak conversion rate max
dTdx
) may decrease
(for example, for the standard PU foam, those Hmax shown in Figure 4.9) with increasing β
and the end result in βHmax increases (MLRmax in Table 4.2). This result is consistent with
the observations of other researchers (Agrawal 1992; Seungdo and Park 1995).
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
101
4.2.3 Thermal dynamic parameters and behaviours
4.2.3.1 Activation energy and pre-exponential factor
The calculation of the pre-exponential factors (A) and the activation energies (E) using the
Coats-Redfern method for the studied materials are listed in Table 4.3. The tested heating
rate was 10 K/min.
Table 4.3 Pre-exponential factors and activation energies for the testing materials
Air 1st 2.6x1015 207.3N2 1st 4.2x1022 268.4
Air 1st 8.9x108 132.8N2 1st 2.9x1011 161.8
Air 1st 6.7x1011 177.4N2 1st 2.4x1020 195.7
Cotton & polyester
Air 1st 2.4x1010 149.3
1st 3.9x1015 210.0
2nd 1.8x103 67.1N2 1st 2.5x1022 287.7
1st 3.7x1010 81.82nd 7.1x1013 87.3
Gas flow Reaction
Stamina PU foam
Air
A (1/s) E (kJ/mol)
Air
Cotton
Pinewood
mountain ash
Standard PU foam
Testing material
The OFW method was also used to check decomposition behaviours under various heating
condition, and compute kinetics. It was found that the slope in the lnβ vs 1/T plot varied, i.e.
two different slopes exist in the lnβ vs 1/T plot, for example in Figure 4.10 for the cotton
fabric. This is a direct evidence to show that different kinetics dominate the decomposition
in these ranges. Therefore two sets of thermal decomposition kinetics can be obtained for
each material studied under different heating rates, and the one from lower heating rate
range are similar to that computed from the Coats-Redfern method. Two reasons may cause
this result: errors in temperature measurement or different decomposition behaviour.
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
102
Figure 4.10 Plotting of lgβ vs 1000/Tmax for the cotton fabric
As discussed in the previous section, the temperature measurement error may exist in tests
of various heating rates. The error in the peak temperature measurement will affect the
accuracy of kinetics, which were obtained by the single running method (as in Equation (4.3)
and (4.4)). In a normally adopted low heating rate range (from 5 to 30 K/min), the sets of
kinetics obtained from different heating rates also varied. Therefore very low heating rate
(10 K/min) was adopted in current study to obtain a set of “ perfect” kinetics.
When the OFW method was used to obtain the kinetics, it calculates the linear relationship
over the heating rates. It was found while the temperature measurement error exists in all
heating rates, it has a minor effect on the slope (activation energy). The kinetics calculated
by the OFW method were mainly determined by the moving trend of the peak parameters.
Therefore, it is believed that OFW method is able to generate more accurate results if a
linear relationship exists.
Different decomposition behaviours of these materials under various heating rates do exist.
As discussed in Section 2.3, there may be a number of reactions in a decomposition process
and several sub-reactions in a reaction. All these reactions and sub-reactions have different
behaviours under various environments and the biggest factor that affects the decomposition
behaviour most is the heating rate. Since the studied materials are all composite, the overall
reaction to describe the decomposition is determined by a combination of individual
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
103
reactions and sub-reactions. Therefore the order and positions of the reactions and sub-
reactions are varied as the heating rate changes. These changes may then affect the kinetics
under different heating rates.
Kinetics calculated from the OFW method for the high heating rate range (50 ~ 200 K/min)
were shown in Table 4.4. From Table 4.3 and Table 4.4, it can be seen that as the heating
rate is increased, the calculated activation energy of the first reaction decreased for all the
materials under both inert and oxidising flow conditions. The tested materials have lower
calculated activation energy under the air flow condition than that under the inert flow. This
matches with the lower peak temperatures under the air flow environment compared to
those of the nitrogen environment, as listed in Table 4.2. Of the multi-reaction materials, the
calculated activation energy of the Standard PU foam for the first reaction drops
dramatically from a low heating rate to a high heating rate and the activation energy of the
second reaction increases gradually. This means the second reaction takes control of the
decomposition process under high heating rates.
Table 4.4 Pre-exponential factors and activation energies for heat rate range (50 ~ 200 K/min)
Air 1st 3.5x108 128.1
N2 1st 9.6x1014 142.3
mountain ash
Air 1st 1.2x108 140.2
Cotton & polyester
Air 1st 5.3x105 95.7
1st 2.1x103 68.3
2nd 5.5x105 95.9
Stamina PU foam
Air 1st 3.5x1010 75.8
Standard PU foam
Air
A (1/s)
Cotton
E (kJ/mol)Testing material
Gas flow Reaction
While no data for identical materials is available from literatures, the following
experimental results from others are still helpful for an approximate comparison. For wood
species the activation energy and the pre-exponential factor of the pyrolysis process are
typically in the order of 102 kJ/mol and 108 1/s, respectively (Moghtaderi 2001). Since
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
104
wood-based fuels are composed of approximately 50 - 60% cellulose by mass, many kinetic
characteristics of cellulose pyrolysis are common to all biomass and wood-based type fuels
according to Antal and Varhegyi (1995). However a wide range of kinetics are reported for
cellulose in a one-step global pyrolysis process. For example, the activation energy ranges
from 70 to 200 kJ/mol and the pre-exponential factor in a range from 6.8×103 to
4.7×1012 1/s, (Barooah and Long 1976; Gullet and Smith 1987; Cordero et al. 1990; Antal
and Varhegyi 1995; Lyon and Janssens 2005). The variation for this single major
component of wood is believed to be caused by the experimental conditions (Vovelle et al.
1984). Some other examples for woods are as follows. Cordero et al. (1991) obtained an E
value of 83 kJ/mol for sawdust of Aleppo pine under a heating range from 5 to 20 K/min.
Atreya suggested an E value of 126 kJ/mol for pine wood (Atreya et al. 1986). For the
standard PU foam, Luo and He (1997) obtained an activation energy of 170 kJ/mol under a
testing condition of 20% O2 flow and 30 K/min heating rate. Similar values were reported
and adopted by a number of other researchers (Chang et al. 1995; Fernando 2000).
Compared with the above experimental data, the test results from the current study, with a
lower heating rate range, fall into the reported ranges with reasonable agreement.
4.2.3.2 Linear relationships and shifting pattern
Three reasonably linear relations were observed among three types of parameters: heating
rate, peak temperature and peak rate in lnβ vs Tmax-1, lnβ vs Hmax
1/2, and Hmax vs Tmax-2 plots.
Figure 4.11 to 4.13 show two of these relations for some of the studied materials, since the
third relation can be derived from the first two. The linear relationships agree with the
previous theoretical analysis, as presented in Equation (4.15) and (4.16).
Notably, the lnβ – Tmax-1 plot, which is used to obtain kinetics such as activation energy, has
the highest linearity in comparison with other two plots. The linear plots for PU foams are
broken into a low heating rate range (10 to 30 K/min, corresponding to lnβ from 2.3 to 3.4)
and a high heating rate range (50 to 200 K/min, corresponding to lnβ from 3.9 to 5.3)
(Figure 4.13). As discussed previously, this is caused by different decomposition
mechanisms on the different heating rate ranges. For the single reaction materials, the
linearity on both heating rate ranges is similar, while the linearity is much higher on the first
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
105
heating rate range for the multi-reaction materials. The linearity for all materials tested in
the current study, varying between 0.975 and 0.998 for the lnβ – Tmax-1 plot, are slightly
lower than that of the rubbers tested by Seungdo and Parks (1995).
1.54
1.56
1.58
1.6
1.62
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
ln (Beta)
1/T
max
x10
00
0.12
0.125
0.13
0.135
0.14
2 2.5 3 3.5ln (Beta)
Hm
ax0.
5
(a) lnβ – Tmax-1 plotting (b) lnβ – Hmax
1/2 plotting
Figure 4.11 Linear relationships for the mountain ash under lower heating rate range
1.5
1.52
1.54
1.56
1.58
1.6
2 2.5 3 3.5ln (Beta)
1/T m
ax x
1000
0.1
0.11
0.12
0.13
0.14
2 2.5 3 3.5ln (Beta)
Hm
ax0.
5
(a) lnβ – Tmax-1 plotting (b) lnβ – Hmax
1/2 plotting
Figure 4.12 Linear relationships for the pine under lower heating rate range
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
106
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1 2 3 4 5 6
ln (Beta)
1/T
max
x10
001st peak
2nd peak
0.08
0.1
0.12
0.14
0.16
1 2 3 4 5 6
ln (Beta)
Hm
ax0.
5
1st peak
2nd peak
(a) lnβ − Tmax-1 plotting (b) lnβ − Hmax
1/2 plotting
Figure 4.13 linear relationships for the standard PU foam under two heating rate ranges
In the current study, these linear relations were observed on a wider heating rate range,
especially at the high heating rate range. This is especially meaningful since the slope of the
linear curve of lnβ – Tmax-1 plotting represents the activation energy. This shifting pattern,
extending the linear relationship to the higher heating rate range, is a clue to the application
of the thermal kinetics obtained from the range to real fire test environment, such as the
cone calorimeter test.
4.2.4 Isothermal test
Isothermal tests for the studied materials were also carried out to investigate the pyrolysis
processes under so-called “ isothermal” (constant temperature) conditions. It is helpful to
investigate the temperature at which pyrolysis starts. This identified temperature can be used
to validate modelling pyrolysis temperatures in Chapter 6.
Shown in Figure 4.14 are mass loss rate curves for mountain ash under various temperature
conditions. Both sample weight and mass loss rate curves for the standard PU foam under
different external temperature conditions are presented in Figure 4.15.
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
107
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
00 2 4 6 8 10 12 14 16 18 20
Time (min)
MLR
(mg/
min
)
250 oC320 oC375 oC
oCoCoC
Figure 4.14 Mass loss rates for mountain ash under different temperature environments
0
10
20
30
40
50
60
70
80
90
100
0 5 10 15 20Time (min)
Sam
ple
wei
ght (
%)
-0.45
-0.4
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
Mas
s lo
ss ra
te (m
g/m
in)
WT-250 oC
WT-300 oC
MLR-250 oC
MLR-300 oC
oCoC
oCoC
Figure 4.15 Isothermal sample weight and mass loss rate curves for the standard PU foam
It can be seen that for the mountain ash, the onset pyrolysis temperature is slightly below
250 oC (in Figure 4.14), since a very low mass loss rate exists at this temperature. It was
seen that a small peak appeared in the first 1 to 2 minutes, which is believed to be caused by
the release of the moisture. When the isothermal temperature increased from 320 oC to
375 oC, the mass loss rate nearly doubled. However the times to the peak mass loss rates
under various temperatures were almost identical.
From Figure 4.15, it was found that the pyrolysis temperature for the standard PU foam
under the isothermal condition is between 250 oC and 300 oC. At 250 oC, the mass loss rate
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
108
was almost zero during the whole testing duration, but there is a significant mass loss at
300 oC. The sample weight change under the 250 oC isothermal condition is resulted from
the release of moisture. The total mass lost during the whole test is relatively small, less
than 20% of the sample mass.
4.3 Summary
To obtain kinetics to describe the pyrolysis process of the studied materials properly,
detailed analysis of the TGA test procedure and computational scheme were performed. A
series of TGA tests were carried out under both dynamic and isothermal conditions. The
data from the tests were used to obtain the desired kinetics in two computing methods. By
analysing the test conditions and linear relationships among the obtained parameters, the
computed kinetics from the higher heating rate range were believed to be suitable to model
decomposition happening in an environment close to real fires.
The Coats-Redfern method was used to calculate “ perfect” kinetics at low heat rate as a
usual scheme. Applying the one-step overall Arrhenius pyrolysis equation, the OFW method
was chosen to compute kinetics through multiple heating rate dynamic tests. The benefits
from the OFW method include higher accuracy and validation ability for the computed
kinetics.
Several factors that may affect the test results were identified. The error in temperature
measuring was found to be controlled by generalizing the test procedure. More importantly
it was found that the calculated kinetics were less sensitive to the peak temperature but
sensitive to the linear slope in the lnβ vs 1/Tmax plotting. The temperature gradient within the
sample was not considered by the author as a major issue, since an even greater gradient
exists in the real decomposition environment. When the usage of these “ effective” kinetics
is to model decomposition in real fire situations, the concern becomes how to obtain the
kinetics in a similar heating condition. Furthermore the kinetics obtained from a very slow
heating rate are of questionable while were uses in the application to a real fire problem.
Through the tests, parameters of the decomposition process under various heating rates
(from 5 to 200 K/min) were obtained. It was noted that the thermal decomposition processes
Chapter 4 Experimental Decomposition Study for the Furnishing Materials ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
109
for the timbers and cotton could be treated approximately as a one-step overall reaction,
while there were two peaks (i.e. two reactions) in decompositions of the PU foams and the
cotton and polyester fabric. These different decomposition characteristics must be taken into
consideration when the pyrolysis model is chosen or when simplifications are made.
Two different kinetics values were obtained by OFW method from two heating rate ranges
(by two different slopes) for the tested materials. By analysing the test procedure and
features of the OFW method, the author believed that there are two different decomposition
behaviours in the lower (10 to 30 K/min) and higher (50 to 200 K/min) heating rate ranges.
The theoretical basis for this judgement was that discussed in previous chapters: that
multiple reactions or sub-reactions have different behaviours under different heating
conditions. The kinetics obtained from the TGA test were all based on a certain degree of
simplification, i.e. how many steps and reactions will be used to describe the decomposition
(Moghtaderi 2001). When a one-step overall pyrolysis model was chosen for the current
modelling, the kinetics obtained from a heating condition, which was closer to that in real
fires, was more applicable. This judgement has to be validated by later modelling and
experimental results.
Linear relationships were observed among a number of parameters that can be found from
either theoretical analysis or test results from others. These linear relationships, especially in
the lnβ vs Tmax-1 plot, which was used to obtain the activation energy, were also found
shifted into that higher heating rate range for the studied materials. This indicates a
possibility of applying the “ effective” kinetics obtained from the higher heating rate range
into an even higher rate, which is closer to that in real fire conditions.
110
mnopqrs!tmnopqrs!tmnopqrs!tmnopqrs!t mHuvJrmwoQxyuMs&zBDr|qrs!=q)~mHuvJrmwoQxyuMs&zBDr|qrs!=q)~mHuvJrmwoQxyuMs&zBDr|qrs!=q)~mHuvJrmwoQxyuMs&zBDr|qrs!=q)~kuMskuMskuMskuMszMvJz%qzuvp\sup\rs|qzBrzMvJz%qzuvp\sup\rs|qzBrzMvJz%qzuvp\sup\rs|qzBrzMvJz%qzuvp\sup\rs|qzBr
As reviewed in Chapter 2, the cone calorimeter test provides an ideal physical platform for
the study of pyrolysis and ignition phenomena. In its confined environment, effects of
various factors, such as radiative flux and gas flow condition, can be well isolated and
identified. Measured parameters from the tests are useful for not only predicting full/real-
scale fire performance but also validating modelling results. Therefore a series of tests were
carried out for the studied materials in the cone calorimeter, and major results are presented
in current chapter.
This chapter includes 5 sections. The first section introduces the principle of oxygen
consumption theory for heat release computation and one of the applications of this theory
in the cone calorimeter. The experiment and calibration procedures are described briefly in
Section 5.2. Major testing results, including ignition time, heat release rate and mass loss
rate, under different radiation levels are presented in Section 5.3. In Section 5.4, the
measured surface temperatures of the specimen are compared with theoretical calculations
as well as published experimental results. Finally, a brief summary of the test results is
given in the last part, Section 5.5.
5.1 Heat Release Rate Measurement
Heat release rate is the primary parameter that contributes to compartment fire hazard from
burning materials. The importance of heat release rate in the fire hazard assessment was first
recognized three decades ago by Smith (1971). At Ohio State University, Smith also
developed one of the first bench-scale heat release rate test methods. As indicated by
Janssens (1995), the importance of the heat release rate can be explained as follows. First,
heat release rate is directly related to mass loss rate. The toxic fire hazard of a material is a
function of the release rate of toxic gases, and the toxic gas release is the product of total
mass loss rate and yield of the gases. Second, the heat released by a material burning in a
compartment results in a temperature rise of the hot layer gases as well as that of
compartment walls and ceiling. The radiation feeding back from the hot gases and surfaces
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
111
to the fuel surface increases the mass loss flux. This means that the heat release will greatly
affect fire development. As illustrated by Babrauskas and Peacock (1992), fire hazard is
most sensitive to changes in the heat release rate of the burning fuel.
In the past half-century, mostly since the 1970s, a number of reaction-to-fire test methods
have been developed to measure the heat release rate. A detailed review for the development
of these methods can be found from one of Janssens’ articles (Janssens 2002). These test
methods vary widely in concept and features. They can be sorted into the following four
catalogues:
Sensible enthalpy rise method (Smith 1972);
Substitution method; first implemented at Factory Mutual Research Corporation,
(Thompson and Cousins 1959);
Compensation method, developed at the National Bureau of Standards (Parker
and Long 1972) and Stanford Research Institute (Martin 1975);
Oxygen compensation method (sourced from Thornton (1917)).
The usage of the first three methods is limited due to their disadvantages. The major
problems in practical implementation of them include temperature delay, thermal lag, and
complexity in controlling. The oxygen compensation method was further improved as the
oxygen consumption method. It is the most accurate and convenient way to measure the net
heat release rate. Problems due to thermal lag are eliminated. It can be applied to both
bench-scale and large-scale tests.
The oxygen consumption method has been proven to be a robust and versatile method with
reasonable accuracy for engineering studies (Huggett 1980). It can be used to study the
dynamic combustion behaviour of flammable materials in controlled or uncontrolled
environments. Therefore the oxygen consumption method was chosen in the current
research.
5.1.1 Oxygen consumption method
The cornerstone of the oxygen consumption method was laid by Thornton (1917). He
indicated that for a large number of organic liquids and gases, a more or less constant net
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
112
amount of heat is released per unit mass of oxygen consumed for complete combustion.
Huggett (1980) found this was also true for organic solids, and further obtained an average
value for the constant of 13.1×103 kJ/kg of oxygen consumed. Thornton’ s observation
implies that it is sufficient to measure the oxygen consumed in a combustion system in order
to determine the net heat released. An equation illustrating the relationship between the
oxygen consumed and heat released was given as:
)( ,''
, pdoxypdairoxyairHPM YmYmEq −= (5.1)
where:
EHPM: heat release per mass unit of oxygen consumed ( ≈ 13.1×103 kJ/kg),
airm : gas flow rate of air (l/s),
pdm : gas flow rate of combustion products (l/s),
airoxyY , : mass fraction of oxygen in the combustion air (0.232 kg/kg in dry air),
pdoxyY , : mass fraction of oxygen in the combustion products (kg/kg).
Some concerns have to be solved before this equation can be applied. For example, the mole
fraction of oxygen that was measured in oxygen analyzers must be converted into a mass
fraction of oxygen in the gas sample. Water vapour needs to be removed from the sample
before it passes through the analyser, so that the resulting mole fraction is on a dry basis.
Such issues have been well addressed and solved in the standard of applying this method
(ISO_5660-1 2002).
5.1.2 Cone calorimeter
An apparatus that utilises the oxygen consumption principle to measure heat release rate is
the cone calorimeter. The cone calorimeter was developed at the National Bureau of
Standards (now renamed as NIST) by Babrauskas in the early 1980s (Babrauskas 1984). It
is presently the most commonly used bench-scale heat release rate measuring apparatus. The
apparatus and test procedure are standardized in the USA (ASTM_E1354-04 2004) and
internationally (ISO_5660-1 2002). The principle construction of a cone calorimeter is
illustrated in Figure 5.1.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
113
Figure 5.1 Illustration of cone calorimeter
A sample material is placed on a load cell which measures the mass in real time. An
electrical cone shape heater provides a uniform external heat source to cause thermal
decomposition and vaporization of the sample material. In the case of piloted ignition tests,
an electric spark is used as the ignition source above the sample. In the case of non-piloted
ignition tests, the ignition of flaming combustion relies on sufficient heating of the sample
and vapor by the cone heater. The product gas mixture is controlled by an exhaust hood and
its flow rate. Temperature and chemical composition of the gaseous product are measured.
The configurations of a standard cone calorimeter are as follows:
Cone heater: consists of a 5 kW electrical heating element, which can provide
uniform radiation up to 100 kW/m2
Spark igniter: an electric spark as the ignition pilot
Specimen holder: holds specimen size of 100×100 mm and can be mounted in
horizontal and vertical orientations
Load cell: used to measure mass changing of the specimen
Oxygen analyzer: high accuracy for oxygen concentration measurement, typical
noise level is 20 ppm
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
114
Air flow: a high temperature duct fan allows flow rate adjusted between 10 and
32 l/s
Additional measurements: most cone calorimeters include instruments for
measuring smoke obscuration and concentration of soot, carbon dioxide, carbon
monoxide, water vapor and other gases.
Depending on the instruments installed, parameters that can be measured as functions of
time from the cone calorimeter may include:
Heat release rate
Total heat released
Mass loss rate
Effective heat of combustion
Ignitability
Smoke and soot yields
Toxic gases yields.
The effective heat of combustion (EHC) is derived from the measured heat release rate per
unit exposed area (kW/m2), "q , and mass flux (i.e. mass loss rate per unit exposed area)
(g/m2⋅s), "m , by the following equation:
""
, mq
H effc ≡∆ (5.2)
Therefore a combustion efficiency, φ , of a material in cone calorimeter testing conditions
can be calculated from following equation:
c
effc
H
H
∆∆
= ,φ (5.3)
where cH∆ is the theoretic heat of combustion.
The cone calorimeter tests for the current research were carried out in the Centre for
Environment Safety and Risk Engineering (CESARE), Victoria University of Technology,
Australia. A Dual Analysis cone calorimeter was used, which was made by Fire Testing
Technology (FTT) Limited, England. The oxygen analyser is a Servomex 1440C Gas
Analyser. More details about this equipment are listed in Table 5.1.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
115
Table 5.1 Specifications of FTT’s cone calorimeter
ITEM SPECIFICATIONS Heater Cone type, 5 kW Heat flux 0 to 100 kW/m2 Specimen size and orientation 100 mm×100 mm, Horizontal & vertical Load cell measuring capacity 500 g, and up to 1300 g for horizontal Load cell tare capacity 2.0 kg Load cell resolution 0.1 g Ignition Electric spark Exhaust fan 0 to 0.045 m3/s Oxygen analyzer detect range 0 to 100%
5.2 Calibration and Operation of Cone Calorimeter
5.2.1 Calibration of cone calorimeter
In the cone calorimeter tests, meticulous operation, routine maintenance, and calibration are
essential for obtaining accurate results. The operation and routine maintenance are described
in the operation manual (FTT 1998).
The calibration procedures include a number of daily calibrations, such as those for gas
analysers, load cell, gas flow rate, heat flux meter, thermocouples, etc. The details of these
calibrations are detailed in the operation manual and standards (AS/NZS_3837 1998; FTT
1998; ISO_5660-1 2002). Among the calibrations, checking for the orifice flow constant,
(calibration constant, named C factor), was carried out daily by a methanol burning. This
check verifies not only the oxygen analyser, but also the integrity of the gas sampling
system, thermocouples, orifice plate and pressure transducer circuits. When a burner is
being supplied with methane (purity ≥99.9%) at a flow rate of 16.76 l, equivalent to 10 kW,
the actual methane flow rate is given in Equation (5.4). Using the basis of 50.0×103 kJ/kg as
the net heat of combustion of methane and the specific constant for methane, 12.54×103
instead of the general constant 13.1×103, basic Equation (5.1) will be expressed as following
equation (5.5), while "q is 10 KW.
=273
3.10176.16
TP
V (5.4)
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
116
2
2
2095.0
)5.1105.1(1025.7 2
O
Oc
X
XP
Tx
C−
−−∆=
−
(5.5)
where:
2OX : the oxygen concentration,
Tg : the gas temperature in the test meter (K),
Tsta: the stack temperature (K),
P : the gas pressure in the test meter (kPa)
∆P: the stack orifice differential pressure (Pa)
V: the flow rate of methane (l/min).
The obtained orifice coefficient should be between 0.040 and 0.046 according to the manual
of the cone calorimeter (FTT, 1998).
Before the normal testing, a burning of a piece of PMMA was also carried out daily under
an external radiation of 50 kW/m2. This test can calibrate heat release rate as well as the
heater radiation, smoke yield and yields of various gases. A typical running result for the
heat release rate of PMMA is shown in Figure 5.2.
0
100
200
300
400
500
600
700
800
0 100 200 300 400 500 600 700 800 900
Time (s)
HR
R (k
W/m
2 )
Figure 5.2 Calibration by PMMA burning
Meanwhile, another calibration method, Heat Release Rate calibration, developed for a
project of interlaboratory-calibration, was also adopted for comparison. In this calibration,
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
117
methanol (99.8% of purity, H2O < 0.1%) was burnt in a set of 4 tests (50 ml, 100 ml, 150 ml,
and 200 ml respectively) to decide the C factor. An initial value of C factor (Ci) was used to
obtain the total heat released in the cone calorimeter tests. Then an amended value of the C
factor can be calculated by following formula (Equation (5.6)).
THRdVCC i /94.19×××= (5.6)
where:
V: volume of methanol,
d: density of methanol (0.791kg/l),
19.94: the net heat of combustion of methanol (MJ/kg), and
THR: calculated total heat release from cone calorimeter tests (MJ/kg).
Average the four results to get final C factor. The standard deviation should be within 2%.
Compared the C factors obtained from methane test and methanol test, the difference is
within 1.5%, which shows a good operation condition of the used cone calorimeter.
5.2.2 Experiment procedure
5.2.2.1 Testing materials and preparation
The furnishing materials in the current cone calorimeter tests are the same as those used in
the TGA tests (Chapter 4). Their properties were listed in Table 4.1.
For the timber materials, two thicknesses, 32 mm and 14 mm, were tested. For the PU
foams, tests were carried out with 50 mm and 25 mm thicknesses. The top area of these
samples was 100×100 mm.
For the wood and PU foam specimens on tests, the side faces and bottom were wrapped
with aluminium foil, and a 13 mm thickness of ceramic fibre was put under foil for thermal
insulation, see Figure 5.3 (a). The fabrics, cut into 120×120 mm, were suspended by an
empty aluminium foil frame, as shown in Figure 5.3 (b). Meanwhile, certain combinations
of the furnishing materials, mainly adding fabrics on the top of the polyurethane foams,
were also tested, following the configuration of a previous test in CSIRO (He et al. 1999).
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
118
(a) wood and foam specimens
(b) fabric specimen
Figure 5.3 Section views of the specimens in the sample holder
Before the tests, all specimens were conditioned in a condition chamber with a temperature
of 23±1 °C and relative humidity of 50±2% for more than 48 hours, according to the ISO
5660-1 standard. Under such conditions, the mass variations for the woods and fabrics are
less than 0.1% per 24 hours, and no mass change were observed for the PU foams.
The moisture content in the timber materials was also measured. After conditioning as
above, they were placed in an oven and heated up to 100 °C. The weight change was
monitored every 24 hours. The change in the average density and the average moisture
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
119
content are shown in Figure 5.4. It can be seen that after 48 hours of heating the density of
the sample became constant and the moisture content for the mountain ash was therefore
measured as 8.5%.
0
100
200
300
400
500
600
700
0 20 40 60 80
Time (hr)
Den
sity
(kg
/m3 )
0
1
2
3
4
5
6
7
8
9
Mo
istu
re R
elea
se (
%)
Density
Moisture Release
Figure 5.4 Moisture content changing curve for mountain ash
5.2.2.2 Testing condition and data sampling
All the materials were tested in the horizontal orientation, with an air flow rate of 24 l/s.
Radiation levels ranged from 10 kW/m2 to 100 kW/m2. Both spark piloted and non-piloted
methods were tested. The exposed areas of the samples were 0.0089 m2 with a standard steel
frame.
The scrubbed method was used for all the tests. In the scrubbed method, all the water and
CO2 content were removed from the sampling gas before it was pumped to the oxygen
analyser. An illustration of the gas flow is given in Figure 5.5, and followed the
requirements discussed in Appendix D of AS/NZS 3837:1998.
In accordance with the test standard for the cone calorimeter, the tests under each condition
were repeated at least 3 times to obtain average values, while the differences of the
measured heat release rates during the 180 seconds from the ignition for the 3 runs were less
than 10%.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
120
Figure 5.5 Illustration of gas sampling in cone calorimeter
The ignition time was determined manually by observing a sustained ignition, rather than a
brief, non-sustained flash. The end of the test was decided by the moment of flame out plus
an extra 2 minutes, according to the ISO-5660 standard. All the other testing parameters
were recorded automatically by a special software, WinCone4.2, which was developed by
the FTT and bundled with the cone calorimeter device.
Tables that listed all the tests conducted are included in Appendix B.
5.3 Experimental Results
5.3.1 Ignition time
5.3.1.1 Tested results
The average ignition times of the tested materials under different radiation levels and
sample thickness are listed in Table 5.2. When estimated accuracy of the time measurement
is approximately within second, the calculated average values were shown with one decimal.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
121
Table 5.2 Ignition time (in seconds) and standard deviations of the tested materials under different conditions
10 75
Yes Yes No Yes No Yes No Yes No No
32 mm NA SM NA153.2 (63.5)
NI53.0 (7.5)
61.3 (6.1)
27.0 (3.2)
36.2 (7.9)
10.7 (2.1)
14 mm NA NA NA98.3
(11.4)NA NA NA
29.3 (2.1)
55.0 (12.5)
NA
Pine wood 32 mm NA SM NA96.3
(27.2)NI
35.0 (1.7)
65.7 (26.0)
19.3 (2.1)
30.5 (5.0)
8.0 (1.0)
Cotton 1 layer NI59.0 (7.0)
NI22.3 (1.5)
28.3 (5.9)
15.3 (1.2)
18.3 (2.5)
11.3 (1.5)
12.0 (1.0)
7.3 (0.6)
Cotton & Polyester 1 layer NI
34.7 (1.2)
NI17.3 (1.5)
18.3 (1.5)
8.7 (1.5)
12.7 (0.6)
6.3 (0.6)
7.7 (2.3)
5.3 (0.6)
50 mm41.3 (3.5)
8.0 (2.6)
NI6.0
(2.0)26.7 (8.4)
4.0 (1.0)
17.0 (5.3)
1.0 (0.0)
10.3 (5.7)
2.7 (0.6)
25 mm NA NA NA7.7
(3.8)46.3
(16.0)NA NA
3.7 (1.5)
4.7 (1.2)
NA
Standard PU foam 50 mm NI
8.0 (3.2)
NI3.3
(0.6)21.0 (5.2)
2.0 (1.0)
8.0 (5.3)
1.7 (0.6)
4.5 (2.4)
1.3 (0.6)
mountain ash
Stamina PU foam
20 30
Piloted Material
Radiation (kW/m2)
Thickness 5040
Notes: 1. The ignition times are average values from at least 3 tests. The standard deviations are listed in brackets.
2. NA: not tested at such condition. 3. NI: no ignition taking place. 4. SM: smouldering and difficult to determine ignition time.
It was noted that under 75 kW/m2 the ignition times of the piloted and non-piloted tests
were almost identical for most of the tested materials. Therefore only the non-piloted results
are presented in the last column of Table 5.2.
5.3.1.2 Linear relationship between 5.0−igt vs radiation
A linear relationship between the tig-0.5 vs radiation for the thermally thick materials has
been reported by Delichatsios (1999). By applying the previous judgment for the thermally
thin or thick materials in Section 2.4.1.1, judgments for the studied materials were
performed. Table 5.3 shows the thermal length characteristic, Lthermal and physical length,
Lphy, for the studied materials.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
122
Table 5.3 Thermal length characteristics for the studied materials Heat flux Ts-TsI Conductivity Lthermal Lphy
(W/m2) (K) (W/mK) (m) (m)timber 300 0.15#1 0.00225 0.032fabric 280 0.1#2 0.0014 0.001
PU foam 300 0.026#3 0.00039 0.05
Material
20000
Note: A relative low radiation was chosen;
The temperature differences for the surface temperatures were determined from ambient to ignition of the materials;
Data source of the thermal conductivities: #1: average from SFPE Handbook (2002)
#2: Lawson and Pinder (2000) #3: Mukaro et al. (2003)
It was found that the timber and PU foams could be considered as thermally thick at various
radiations, since their thermal length characteristics are much smaller than their physical
length (thickness). On the contrary, the fabrics can be considered as thermally thin due to
their larger thermal length characteristics compared to their physical length. The estimation
result for the fabrics matches the experimental results carried out by Flesischmann and Chen
(2001). They investigated 14 different fabrics, including 100% cotton and 51% polyester
and 49% cotton, and found that above 15 kW/m2 radiative flux, a good correlation existed
between 1/tig versus evq ′′ , which is observed for thermally thin materials. Therefore all the
fabrics should be considered as thermally thin.
From the current tests, a linear relation was found between 1/tig1/2 and evq ′′ for the timber and
PU foams (see Figure 5.6) since they were all thermally thick materials based on the above
calculations. It was interesting to note that the same trend was also seen for the fabrics, the
cotton and the cotton and polyester, although they were normally treated as thermally thin
materials. This may be explained by the configuration of fabric samples in the tests. The
fabrics were suspended over a volume full of air which is a poor heat conductor with a
thermal conductivity of 0.025 W/m⋅K. Therefore, there was little heat lost through the back
side, or the empty side, of the sample. The thin fabrics then behaved like a thermally thick
material.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
123
y = 0.009x + 0.2639 R2 = 0.8677
y = 0.0049x + 0.1076 R2 = 0.8663
y = 0.0055x - 0.0561 R2 = 0.9955
y = 0.005x - 0.0627 R2 = 0.9966
y = 0.0042x + 0.074 R2 = 0.9501
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0 10 20 30 40 50 60 70 80
External Radiation (kW/m2)
1/t ig
1/2 (1
/s1/
2 )
mountain ash
Pine
Cotton and polyester
Standard PU foam
Cotton
Figure 5.6 Relation between ignition time and external radiation
5.3.1.3 Determining the critical radiative flux
The critical radiative flux, which was defined in Chapter 2 as the minimum radiative flux
required for ignition to occur, was not measured in the current experiments. However it can
be determined approximately by a method, which was developed by Quintiere (Quintiere
and Harkleroad 1984) and discussed in Chapter 2, Figure 2.4. A Quintiere chart of the
ignition time applied for the mountain ash is shown in Figure 5.7.
0
100
200
300
400
500
600
0 10 20 30 40 50 60 70 80
External Radiation (kW/m2)
Igni
tion
time
(s)
Trendline
Predicted critical radiative flux
Figure 5.7 Ignition times and trend-line for the mountain ash
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
124
By extending the trendline of the curve of ignition time back to 600 seconds, the critical
radiative flux for the mountain ash was then estimated approximately as 17 kW/m2. This
critical radiative flux can be used to predict ignition surface temperature, as discussed in
Section 2.4.2.
This estimated critical radiative flux could be validated indirectly by the following methods.
The first one is observation from the tests. The smouldering combustion always occurred at
20 kW/m2 radiation but not at 10 kW/m2 for the mountain ash. Secondly it can be validated
by an empirical equation given by Babrauskas (2001), which was derived from a great
number of cone calorimeter tests for wood. The details of this comparison can be found later
in Section 5.3.1.5.
5.3.1.4 Effect of sample thickness
It was found that for the mountain ash, varying the sample thickness had different effects on
the ignition time under different radiation fluxes. The ignition time becomes shorter as the
sample thickness (as well as the sample mass) decreases under low radiation flux (<30
kW/m2). On the contrary, under high external radiation, the ignition times for various
thicknesses were similar (piloted ignition) or the ignition time for a thinner sample became
longer (non-piloted ignition).
Ignition of solid fuel is greatly dependent on the sample surface temperature. Under low
external radiation conditions, the surface temperature rise is controlled by conduction as
well as external radiation. The time to heat a thinner sample is shorter than that for a thicker
one. Therefore the thinner sample reaches a certain surface temperature quicker which
results in a shorter ignition time. Under high external radiation conditions, the heat flux
wave within the solid reaches to the back of sample much quicker and bounces back due to
the effect of the insulation material. This bounced back heat wave may speed up the forming
of a thicker char layer. As discussed in Section 2.3.3, the thicker char functions to impede
the heat flux flow in and the gas pyrolysis products out of the solid. Therefore a longer
ignition time may exist where thickness is reduced. This phenomenon can be also found
from experimental results by others (Harada 2001).
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
125
5.3.1.5 Comparison with results from others
For the pine wood, the measured results are similar to Delichatsios’ cone calorimeter
experimental results (Delichatsios 1999). In his piloted experiments, the ignition times for a
19 mm thick pine wood are 132 s and 22 s under 30 kW/m2 and 50 kW/m2 radiation levels
respectively.
The tested results can be validated against even wider experimental data by a correlation
developed by Babrauskas (2001). For the piloted ignition, an estimated rule for radiant
heating ignition of wood is:
( ) 82.1""
73.0130
crev
igqq
t−
= ρ (5.7)
where:
tig: ignition time, s;
evq ′′ : effective radiation flux, kW/m2;
crq ′′ : critical radiation flux, kW/m2;
: sample density, kg/m3.
Checking the current tested results against Equation (5.7), a quite good agreement was
obtained. For example, applying a critical radiation flux of 17 kW/m2 obtained from Figure
5.7 for the mountain ash, the measured ignition times from Table 5.2 for the timbers match
the above relationship reasonably well, as shown in Table 5.4.
Table 5.4 Comparison of ignition times from current tests and Babrauskas’ equation for mountain ash
30 50 75Tested 153.2 27.0 10.7Estimated 142.9 26.2 9.4
Radiative flux (kW/m2)
Ignition time (s)
5.3.2 Heat release
In this section, experimental results for two types of heat release parameters, the heat release
rate (HRR) and total heat released (THR) are presented.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
126
5.3.2.1 Heat release rate
The heat release rate curves for the various tested materials under various test conditions are
shown in Figure 5.8 to Figure 5.12. The presented curves were picked up from one of those
3 runs that had a closest peak value to the averaged peak value.
0
50
100
150
200
250
300
350
400
0 500 1000 1500 2000 2500 3000 3500 4000
Time (s)
Hea
t re
leas
e ra
te (
kW/m
2 ) 30kW/m2
50kW/m2
75kW/m2
Figure 5.8 Heat release rate of mountain ash from piloted tests
0
50
100
150
200
250
300
350
400
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Time (s)
Hea
t rel
ease
rat
e (k
W/m
2 )
20kW/m2
30kW/m2
40kW/m2
50kW/m2
Figure 5.9 Heat release rate of cotton and polyester from piloted tests
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
127
0
50
100
150
200
250
300
350
400
450
0 50 100 150 200 250 300
Time (s)
Hea
t re
leas
e ra
te (
kW/m
2 ) 10kW/m2*20kW/m2
30kW/m2
40kW/m2
50kW/m2
Figure 5.10 Heat release rate of Stamina PU foam from piloted tests
(* smouldering and flame-out during the tests)
0
50
100
150
200
250
0 500 1000 1500 2000 2500 3000 3500
Time (s)
Hea
t rel
ease
rat
e (k
W/m
2 )
Non-pilot
Pilot
Figure 5.11 Piloted and non-piloted HRR curves for mountain ash under 50 kW/m2
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
128
0
20
40
60
80
100
120
140
160
180
200
0 500 1000 1500 2000 2500 3000 3500 4000
Time (s)
Hea
t rel
ease
rat
e (k
W/m
2 )
30kW/m2(32mm)30kW/m2(14mm)50kW/m2(32mm)50kW/m2(14mm)
Figure 5.12 Effect of different thicknesses on HRR for mountain ash
There are two obvious peaks shown in the mountain ash’ s curves (Figure 5.8). The first
peak corresponds to combustion of all the available combustible gas mixture, which was
generated before the ignition. After the ignition occurred, the top-charring layer becomes
thicker and thicker, which weakens the heat transfer downward and obstructs the pyrolysate
from escaping upward. As discussed in Chapter 2 and the previous section, for a thermally
thick material, the inside temperature of the sample increases rapidly due to the “ bounced
back” heat front. When the inside temperature reaches above the vaporization temperature
of water and the pyrolysis temperature of the timber, the vaporization and pyrolysis
processes occur within the sample. The generating of those gasified products results in high
pressure building up inside which will cause cracking. The exposure of inner virgin material
contributes to the forming of the second peak in HRR curves. This phenomenon was
previously observed and explained by a number of researchers (Tran 1992; Grexa et al.
1997; Harada 2001). It was also proved by these researchers through tests without an
insulation material present. In that type of tests, the second peak did not appear since the
sample holder behaved as a heat sink and no high pressure was built up within the sample.
For the fabric materials, the peak value of the heat release rate increases (and needs less
time to reach the peak) when the external radiation is increased (Figure 5.9). However, the
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
129
total heat release involved in the combustion period under various radiation conditions are
very similar except for the lowest radiation level, 20 kW/m2, see Figure 5.13.
For the polyurethane foams, the peak HRR and the time to reach the peak are almost
independent of the external radiation heat flux unless it is below 20 kW/m2. The lowest
radiation used for the polyurethane foam was 10 kW/m2, and this is obviously close to the
critical radiation flux. All three tests under this radiation flamed out (extinguished) during
the tests. Therefore a lower peak HRR and a shorter combustion period were observed for
this testing condition (Figure 5.10).
Excluding the period leading to ignition, the curve shapes for both piloted and non-piloted
tests are very similar for each of the tested materials. An example is given in Figure 5.11 for
the mountain ash. Under high radiation levels, the two curves obtained from piloted and
non-piloted combustions were almost identical.
The effect of sample thickness on the HRR curve shapes and peak values depends on the
external radiation conditions, as shown in Figure 5.12. Under high radiation, varying of
thickness has less effect on the value of the first peak in the HRR curve. Under low
radiation, the thinner sample has a higher value for the first peak. As explained previously,
the first peak is greatly determined by the initial heating conditions. Under high radiation,
the sample was heated rapidly and the time for the first peak to occur is relatively short. The
conductive heat losses through the samples are relatively small and the difference between
the thicknesses is also small. Therefore similar surface temperature profiles on various
thickness samples result in similar first heat release peak behaviour (value and time
occurred).
Under lower radiation, the role of heating of the samples is obviously slower. Over the
longer heating period before the ignition, the difference of heating effort for different
sample thickness is significant. The thinner sample will reach a higher surface temperature
much more quickly due to its smaller mass.
As for the second peak, the thinner samples reach the “ cracking” condition (as discussed
previously) more quickly. Therefore the times to the second peak were reduced significantly
under various radiation condition levels.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
130
5.3.2.2 Total heat released
Total heat released (THR) is another key parameter. Shown in Figure 5.13 are examples of
the total heat release histories of the cotton and polyester under various radiation levels. It
can be seen that the total heat released during combustion is virtually identical except under
the lowest radiation condition. Under the lowest radiation the combustion efficiency is
relatively low which will be shown in next sub-section.
0
1
2
3
4
5
6
7
8
0 50 100 150Time (s)
Tota
l hea
t rel
ease
(MJ/
m2 )
20kW/m2
30kW/m2
40kW/m2
50kW/m2
Figure 5.13 THR of the cotton and polyester under various radiation conditions
The total heat release from each cone calorimeter test can be found in Appendix B.
5.3.3 Mass loss rate and effective heat of combustion
The mass loss of a specimen is a direct measure of its pyrolysis. Figure 5.14 to 5.16 show
the mass loss rate (MLR) and mass flux curves for the wood, PU foam and fabric under
different radiation levels. The shapes of the MLR curve for the thermally thick materials are
similar to the corresponding heat release rate curves, shown from Figure 5.8 and Figure 5.12.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
131
0
2
4
6
8
10
12
400 820 1240 1660 2080 2500 2920 3340 3760 4180
Time (s)
Mas
s fl
ux
(g/m
2 s)
0
0.02
0.04
0.06
0.08
0.1
Mas
s lo
ss r
ate
(g/s
)
Figure 5.14 Mass loss of mountain ash under 50 kW/m2 piloted
0
2
4
6
8
10
12
14
16
18
0 50 100 150 200 250
Time (s)
Mas
s fl
ux
(g/m
2 s)
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Mas
s lo
ss r
ate
(g/s
)
Figure 5.15 Mass loss of Stamina PU foam under 50 kW/m2 piloted
0
2
4
6
8
10
12
14
16
18
0 10 20 30 40 50 60 70 80 90 100 110
Time (s)
Mas
s flu
x (g
/m2 s
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Mas
s lo
ss r
ate
(g/s
)
Figure 5.16 Mass loss of the cotton and polyester under 50 kW/m2 piloted
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
132
As described earlier in this chapter, the effective heat of combustion can be calculated from
HRR and MLR data from Equation (5.2). Figure 5.17 shows an example for the mountain
ash under 50 kW/m2.
The combustion efficiency can be approximately estimated by Equation (5.3). The
theoretical heat of combustion can adopt an average value, 20 MJ/kg, which is suggested by
the SFPE Handbook for normal timber materials. Then the combustion efficiency is
computed. The EHC measured varies from test to test as well as at different stages of a
single burning. For the test shown in Figure 5.17 the combustion efficiency is estimated
approximately 50% for most of the testing period. The first peak that exceeds 20 MJ/kg may
be caused by combustion of accumulated gas phase pyrolysis products and then it does not
correspond exactly to the mass loss rate at that moment.
0
6
12
18
24
0 400 800 1200 1600 2000 2400
Time (s)
EH
C (
MJ/
kg)
Figure 5.17 Effective heat of combustion for the mountain ash under 50 kW/m2
It was also found that for the tested timber materials a small peak exists in the MLR curves
before ignition occurred except for that under very high radiation levels (> 75 kW/m2).
Figure 5.18 provides some examples, which show partial enlarged MLR curves for the
mountain ash under various radiation levels. In this figure, the small peak can be observed
in the MLR curves under 30 and 50 kW/m2 radiation levels.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
133
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0 10 20 30 40 50 60 70 80 90Time (s)
MLR
(g/s
)
30 kW/m2 (tig=77s)
50 kW/m2 (tig=23s)
75kW/m2 (tig=11s)
Figure 5.18 Small peaks in the MLR curves for mountain ash
under various radiation levels
This phenomenon has been reported by other researchers. For example, similar results can
be found from cone calorimeter test data of the Worcester Polytechnic Institute (WPI) for a
wide range of wood-based materials (WPI). From their results it is known that for a very
thin sample (3 mm in their tests) this phenomenon would not happen under various radiation
levels.
This can be explained by the basic charring process and Atreya’ s theoretical analysis that
were presented in Chapter 2 and 3. It is known that the mass loss before ignition was
affected by the release of both moisture and pyrolysis products, including inert gases. When
a thick char layer formed before ignition under lower radiation levels, it reduced the release
rates by obstructing the heat and mass transfers. Then the mass loss rate curves drop until
the pyrolysis generating and moisture releasing rates increased again by a higher surface
temperature. Under very high radiation levels or for a very thin wooden sample, the
possibility for the thicker char forming before the ignition is quite low. Therefore no such
obvious peak in the MLR curve before ignition can be observed. Watt gave a similar
explanation for this phenomenon (Watt et al. 2001).
To investigate effort of the moisture in the mass loss rate, some tests with “ dried” timber
samples were conducted. The timber samples are mountain ash which were dried in an oven
with 100 oC temperature setting for over 100 hours. The moisture release (weight loss) is
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
134
approximate 9% and stable sample mass was observed after 72 hours. Show in Figure 5.19
are MLR curves from tests of those “ dried” mountain ash compared with result from normal
conditioned sample.
Figure 5.19 Comparison of MLRs from moisture and dried mountain ashes under 30kW/m2
It’ s noticed that earlier ignition was achieved from those tests of the “ dried” samples. The
small peak value of mass loss rate before the ignition still exists but with lower value. This
lower mass loss rate consists pyrolysis products, including inert gases. It’ s also noted that
moisture release exists in the whole heating process even for the “ dried” samples since total
water content, including free and bound waters, in timber can reach 60% of total mass,
according to discussion in Section 2.3.3.2. When moisture in upper layer was evaporated at
the early heating stage, the moisture inside the sample specimen was keeping heated and
evaporated towards both sides of the sample as the heating front moving in. Converting
values from above graph into mass flux, approximate 0.009 m2 top surface area of the
sample, the moisture release before the ignition from the “ moisture” samples can reach a
rate of couple of grams per square metre.
5.3.4 Species yield
The CO and CO2 yields were also measured in the cone calorimeter experiments. The
measured results for these parameters greatly depend on the determination of end points for
data collection and calculation. So far no universally accepted method for data treatment is
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
135
available in the literature. From previous TGA experiments and theoretical analysis, it is
found that the fraction of conversion of sample mass at maximum conversion rates is almost
constant. This has been adopted to develop a special OFW method to compute kinetics in
Section 4.1.2.2. In the current cone calorimeter experiment, a similar trend was also
observed for the fraction of sample mass, although the fractions varied with the level of
external radiation. For example, the fractions of sample mass consumed for the second peak
heat released for the mountain ash under 30 kW/m2 and 50 kW/m2 are about 67% and 66%
respectively. Shortly after that point (from a point of mass fraction of 72%) the yields of CO
and CO2 keep increasing sharply. The CO yield information associated to the peak HRRs
have been submerged by this sharp increasing. An example of CO yield for the mountain
ash under 50 kW/m2 was shown in Figure 5.20. Since this sharp increase of CO and CO2
yields may be caused by incomplete combustion as well as noise of mass measurement
under a very low mass loss rate, the peak yield values for the mountain ash were only
calculated until 70% of the sample was consumed. These data are shown in Table 5.5.
0
20
40
60
80
100
0 400 800 1200 1600 2000 2400 2800 3200Time (s)
Sam
ple
wei
ght (
%)
0.00
0.04
0.08
0.12
0.16
0.20
CO
Yoe
ld (k
g/kg
)
Sample Weight (%)CO Yield
Figure 5.20 Relation between mass loss and CO yield for the mountain ash
For the average yields, avespecY , , the data generated by the cone calorimeter software is
evaluated using the following formula:
n
m
m
n
YY
n
i isamp
ispecn
iispec
avespec
== ∆
∆
== 1 ,
,
1,
,
(5.8)
where:
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
136
avespecY , : average yield of a species (kg/kg);
ispecY , : measured species yield at ith time step (kg/kg);
n: number of measuring point;
ispecm ,∆ : incremental of a species production rate at the measuring point i, (kg/s);
isampm ,∆ : incremental of the sample mass loss rate at the measuring point i (kg/s).
It can be seen from Equation (5.8) that the accuracy of the average yield of species may be
affected by both accuracy of the gas analyzer ( ispecm ,∆ ) and the noise of each measurement
step of sample mass ( isampm ,∆ ). For example, the average CO yields for the mountain ash
under 30 kW/m2 and 50 kW/m2 radiation levels are 0.023 and 0.014 kg/kg respectively.
These values are much higher than those reported by other researchers which is shown in
the later part of this section. Therefore another improved method was introduced by the
author as in the following equation.
M
dtmY spec
t
avespec
end 0,
= (5.9)
where:
specm : species generating rate (kg/s);
tend: end of the period for sampling (s);
M: total mass loss of the sample (kg) .
The total mass loss, M, can be measured accurately at the end of a test.
An expression for the specY at any given time can be given. Equation (5.9) is a special case.
Now the accuracy of the average yields only depends on the resolution of the gas analysers.
The effect of the mass measuring noise is eliminated. The average yields of CO and CO2 for
the mountain ash, stamina PU foam and cotton and polyester are listed in Table 5.5. Figure
5.21 shows a comparison of these two different gas curves, gas production rate and gas yield,
for the mountain ash under 30 kW/m2. Therefore, in the later part of this section, the gas
yields from the combustion are provided by the gas production rate.
It can be seen from Table 5.5 that average yields of CO and CO2 from the fabric and
polyurethane foam are quite stable under various radiation and pilot conditions. For the
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
137
timber material, a similar trend exists under the piloted ignition condition. However, a
higher radiation level produces much higher combustion efficiency and more complete
combustion which generates less CO and more CO2.
Table 5.5 Yields of CO and CO2 (kg/kg) for testing materials
NI: no ignition took place.
0.0
0.2
0.4
0.6
0.8
1.0
0 700 1400 2100 2800 3500 4200Time (s)
CO
yie
ld (k
g/kg
)
0.000
0.001
0.002
0.003
0.004
0.005
CO
pro
duct
ion
rate
(g/s
)
COYCOP
Figure 5.21 Comparison of CO yield and production curves for mountain ash
(piloted combustion under 30 kW/m2)
Piloted Non-piloted Piloted Non-
pilotedMax 0.048 NI 0.0079 0.0059Aver. 0.0104 NI 0.0034 0.0028Max 1.46 NI 1.75 1.76Aver. 1.036 NI 1.15 1.17
CO Aver. 0.043 0.05 0.045 0.049CO2 Aver. 1.87 1.86 1.87 1.88CO Aver. 0.025 0.026 0.031 0.029CO2 Aver. 2.05 2.07 1.99 2.03
Stamina PU foam
mountain ash (70%
consumed)
At 50 kW/m2
radiation
CO
CO2
Cotton & Polyester
Material Gas Value type
At 30 kW/m2
radiation
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
138
The measured CO yield results matched other experimental data reasonably well when the
improved method was adopted. For example, Babrauskas (1992) provided typical average
CO yield test data from various cone calorimeter experiments as follows:
Douglas fir: 0.003 to 0.005 (35 to 75 kW/m2)
Rigid PU foam: 0.04 to 0.08 (35 to 75 kW/m2).
It can be seen that with one exception (the CO yield of the 30 kW/m2 test, which is slightly
higher), all other tested data matched with the results provided by Babrauskas quite well.
The full histories of CO and CO2 production rates for the cotton and polyester, Stamina PU
foam and pine are shown in Figure 5.22 to 5.24 respectively.
By studying the gas release rate curves, it was found that CO2 release rates have similar
shapes as that of HRR curves. This can be explained because CO2 is a complete combustion
product and is greatly determined by the fuel burning rate, which is normally measured by
the HRR. As for the CO production rate, the latter peak value reflects a great degree of
incomplete combustion, which is quite normal at the final stage of timber combustions.
Number test results showing a similar feature can be found from the cone calorimeter tests
carried out by WPI for various timber based materials (WPI).
0
0.1
0.2
0.3
0.4
0.5
0 10 20 30 40 50 60 70 80 90 100 110 120Time (s)
CO
2 re
leas
e ra
te (g
/s)
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
CO
rel
ease
rat
e (g
/s)
CO2 (30kW/m2)
CO2 (50kW/m2)
CO (30kW/m2)CO (50kW/m2)
Figure 5.22 Gas release rates from the cotton and polyester
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
139
0
0.1
0.2
0.3
0.4
0.5
0 25 50 75 100 125 150 175 200 225 250
Time (s)
CO
2 re
leas
e ra
te (g
/s)
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
CO
rel
ease
rat
e (g
/s)
CO2 (30kW/m2)
CO2 (50kW/m2)
CO (30kW/m2)CO (50kW/m2)
Figure 5.23 Gas release rates from the Stamina PU foam
0
0.05
0.1
0.15
0.2
0 500 1000 1500 2000 2500
Time (s)
C0 2
rel
ease
rat
e (g
/s)
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
CO
rel
ease
rat
e (g
/s)
CO2 (30kW/m2)
CO2 (50kW/m2)
CO (30kW/m2)CO (50kW/m2)
Figure 5.24 Gas release rates from the pine
5.4 Surface Temperature Measurement
5.4.1 Surface temperature
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
140
As discussed in Section 2.4.3.1, the surface temperature, especially the temperature at which
ignition occurred, is a very important parameter, which has been adopted as a criterion for
ignition. Therefore, in the current cone calorimeter tests, the surface temperature of some of
the materials was measured.
For woods and fabrics, this surface temperature was measured by using 0.1 mm
thermocouples. The configuration of the installation of the thermocouple is shown in Figure
5.25 for the woods. The thermocouple was pushed into the sample along the texture to
obtain good contact. Variation in the installation may result in a tiny gap between the
sample surface and the top of the thermocouple. This gap (d) was estimated to be within
0.1 mm. For the fabrics, the thermocouples were sewn into the texture to obtain good
contact.
Figure 5.25 Installation of thermocouple on the surface of wood
Some surface temperature measuring results are shown in Figure 5.26 and Figure 5.27. Note
that a number of tests were conducted for each test condition and the temperature curves
presented in these figures are the results of the tests in which ignition times best matched
with average ignition time.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
141
0
100
200
300
400
500
0 10 20 30 40 50 60 70 80Time (s)
Sur
face
tem
pera
ture
(oC
)
30kW/m2 (tig=74s)
50kW/m2 (tig=29s)
75kW/m2 (tig=11s)
100kW/m2 (tig=6s)
Figure 5.26 Surface temperatures of the mountain ash
0
50
100
150
200
250
300
350
400
450
500
550
0 5 10 15 20
Time (s)
Sur
face
tem
pera
ture
(oC
)
C&P: cotton&polyester; C&P+PUa: cotton&polyester on top of standard PU foam
C&P 30kW/m2 (tig=12s)
C&P 50kW/m2 (tig=6s)
C&P+PUa 30kW/m2 (tig=14s)
Figure 5.27 Surface temperature of the cotton and polyester and the standard PU foam
Due to the different thermal radiation absorption coefficients of the thermocouple and the
samples, the thermocouple temperature reading may be slightly different from the
temperature of the sample. The thermocouples were pressed into the timber specimens
without additional fixing. The thermocouples usually detach from the surface of the samples
shortly after ignition takes place.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
142
The error in surface temperature measurement was not able known exactly. However it can
be estimated from following aspects. First, the limit of error of used type of theromocouples
(K type from Omega Engineering Inc.) is approximately 2.2 oC or 0.75% of measuring
range. So the error from thermocouple itself is below than 4.0 oC before ignition (<400 oC)
of the tested timber materials. Second part of the error comes from the different emissivities
of timber and thermocouple. According to research by Urbas and Parker (2005), the latter
part of error can be eliminated by maintaining a good contact between the thermocouple and
specimen. Results from another two non-contact surface temperature measuring
technologies are normally adopted for comparison purpose. One is using infrared pyrometer,
which has error limit in an order of 10K within 1200 K range. The other is by
thermographic phosphor, which has an accuracy of the order 1 ~ 5 oC. By checking
temperature profiles from the various technologies, it’ s noticed that when the good contact
was secured which is the case before ignition, temperature measured from thermocouple for
timber materials matches with results from other two techniques quite well. Reported
temperature difference is within 10 oC (Omrane et al. 2003; Urbas et al. 2004). Therefore
the total error of surface temperature measured by thermocouple can be estimated within 10
~ 20 oC.
The temperature readings from the thermocouples became inaccurate after ignition took
place, due to the detachment of the thermocouple from the surface and enhanced radiation
contributed by the flame. The measured surface temperatures at the moment of ignition for
the mountain ash are shown in Figure 5.26. When the radiation flux drops from 100 kW/m2
to 30 kW/m2 the measured surface temperature at the time of ignition increased from
approximate 290 to 320 oC. These values and the trend match experimental data from other
researchers. For example, Yang et al. (2003) measured surface temperatures of 290, 300 and
310 oC for radiations of 50, 40 and 30 kW/m2 respectively for beech wood. Tzeng et al.
(1990) found that the surface temperatures at ignition for a wide range of woods, maple, red
oak, poplar and mahogany, are quite close within a temperature range from 580 K to 660 K
(307 to 387 oC) under external radiation from 21 kW/m2 to 36 kW/m2. A trend that surface
temperature at ignition decreases with the increase of the external radiation level was found
for all the tested materials, such as shown in Figure 5.26 and Figure 5.27. This trend is
different from some reports but in agreement with others (Moghtaderi et al. 1997; Bhargava
et al. 2001).
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
143
In the results for the fabrics (Figure 5.27) another configuration was added, i.e. fabric was
placed on top of the PU foam. Comparing the results from the single fabric configuration
and the combination of fabric and PU foam, similar temperature rising trends were observed.
With the PU foam covered by the fabric, the surface temperature of the fabric was about
20 °C lower due to the heat transfer to the PU foam.
The possible temperature difference caused by various installation depths of the
thermocouple was investigated. Figure 5.28 shows the temperatures at different depths from
a test on the mountain ash under 40 kW/m2 radiation. From the results shown, it can be
estimated that the temperature measuring error caused by the different depths (not more than
0.1 mm) will be within 10 to 20 °C under the test condition.
As for the possible difference between the temperature measured by the thermocouple and
the sample surface temperature, Urbas and Parker (1993) carried out an investigation on
Douglas fir in a cone calorimeter environment. By comparing results from thermocouple
and infra-red pyrometer measurements, they found that an accurate measurement can be
achieved when a good contact between the thermocouple and the surface maintained, and
the thermocouple was not located in the proximity of a fissure.
0
100
200
300
400
500
600
0 5 10 15 20 25 30 35 40
Time (s)
Su
rfac
e te
mp
erat
ure
(oC
) Depth=0.1mm
Depth=0.5mm
Figure 5.28 Surface temperatures of the mountain ash at different depths under 40 kW/m2
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
144
The average surface temperatures at ignition and standard deviation for the mountain ash is
given in Table 5.6. In the table average values of critical mass flux and its deviation are also
shown.
Table 5.6 Average critical surface temperature and mass flux for mountain ash
External Radiation (kW/m2)
Thickness (mm)
Tig
(oC)STDEV of Tig
Mass Fluxcr
(g/m2s)
STDEV of Mass Fluxcr
30 32 330 2.5 6 0.00450 32 334 5.1 8 0.00575 32 290 10
5.4.2 Temperature rising rate
Temperature rising rate is an important parameter since it determine the heating condition
greatly. As discussed in Chapter 2, different heating conditions may cause different
pyrolysis behaviours. Shown in Figure 5.29 and Figure 5.30 are rising rates of the surface
temperatures for some tested materials are. These temperature rising rates are to be
compared with the heating rates in the TGA tests.
0
200
400
600
800
1000
0 10 20 30 40 50 60 70Time (s)
Ris
ing
Rat
e in
(K/m
in)
0
4
8
12
16
20
Ris
ing
Rat
e of
Sur
face
Tem
p. (
K/s
)
mountain AshPine
Figure 5.29 Surface temperature rising rates for the timbers under 30 kW/m2
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
145
0
10
20
30
40
50
0 5 10 15 20Time (s)
Ris
ing
Rat
e o
f Su
rfac
e T
emp
. (K
/s)
0
600
1200
1800
2400
3000
Ris
ing
Rat
e (in
K/m
in)
Cotton&polyester
Cotton
Figure 5.30 Surface temperature rising rates for the fabrics under 30 kW/m2
It was found that the average rate of temperature rise for timber materials under 30 kW/m2
radiation is close to the maximum heating rate in the TGA equipment used previously. The
temperature rising rates for wood under high radiation (50 kW/m2), or for fabrics under all
radiation conditions, are several times higher than the maximum heating capacity of the
TGA. This indicates a necessity for applying the kinetics obtained from the high heating rate
TGA tests and further extending the linear shift pattern to higher heating rates.
5.5 Summary of the Cone Calorimeter Test Results
The cone calorimeter tests were carried out for the studied materials with external radiative
flux ranging from 10 to 100 kW/m2. Major parameters, including ignition time, heat release
rate, mass loss rate and surface temperature, were measured. These parameters provided key
information for describing the ignition and combustion phenomena in the cone calorimeter
environment and are useful for predicting full/real-scale fire behaviour. More importantly,
as discussed at the beginning of this chapter, the test provided a physical platform for the
models to model on and target results to follow. In other words, these parameters are critical
for the modelling of the thermal decomposition process as well as the improvement of
ignition criteria.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
146
The ignition time was measured under various radiation levels and ignition assisted methods
for all materials. Normally the ignition times under non-piloted conditions are longer than
that of piloted conditions. When radiative flux increased up to 75 kW/m2, the difference in
the ignition times between piloted and non-piloted methods for all the testing materials
becomes too small to identify.
Meanwhile an interesting phenomenon between the ignition time and sample thickness was
noticed. A longer ignition time was found for a thinner mountain ash sample under the
50 kW/m2 radiation level. This is believed to be caused by a thicker char layer formed in the
thinner sample which obstructs the release of pyrolysis gases. Even though this phenomenon
only exists in a few cases, it is worthy of further experiment and modelling.
As a major targeting parameter for modelling research of ignition, the variation from the
tests for the ignition time will become a bigger issue. It was found that the non-piloted tests
had a higher variation for ignition times. Meanwhile the variations of the ignition time
measurement increased when the radiative flux decreased, for all the tested materials. The
high variation of the ignition time is usually associated with high variation in other
measured parameters, such as HRR and mass loss rate. It is expected that predicting the
ignition time under lower radiation levels and non-pilot conditions will be a challenge for
any ignition and combustion model.
A critical radiative flux was derived from the ignition time data by the Quintiere chart for
the mountain ash. This critical radiation, besides being able to predict ignition surface
temperature by the equation developed by Quintiere, was also used to validate the ignition
times via the empirical equation suggested by Babrauskas. The predicted ignition times
matched reasonably well with the measurement from the current tests.
Another major output from the cone calorimeter, the HRR curves, was also measured. The
repeatability of the HRR curves, which can be verified by the raw data presented in
Appendix B, met with requirements of the ISO-5660 standard. For the timber materials
there are two peaks in the HRR curves. The latter resulted from the exposure of inner virgin
materials to the radiative flux. Comparing the piloted and non-piloted methods, they have
similar HRR curves except that a higher first peak value existed in the piloted test condition.
Chapter 5 Cone Calorimeter Study for Ignition Properties ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
147
The mass loss rate curves have similar shapes with the HRR curves from the same test for
the tested materials. It was found that there is a small peak before the ignition in MLR
curves under certain radiation levels for the timber materials. By conducting tests with
“ dried” specimen, composition of the mass flux was verified consisting pyrolysis products
and moisture before the ignition. This local peak is formed by effect of impeding mass
release by a gradually formed thicker char layer. When the radiative flux reaches up to 75
kW/m2, this peak disappears due to the faster heating process.
The species yields for CO and CO2 were also monitored. When the CO2 has a similar curve
shape with that of HRR and MLR, the CO curve is quite distinctive from other curves.
Normally multiple peaks can be found in the CO release curve. A data processing technique
was improved to eliminate the errors in average species yields measurement. This error was
introduced by mass measurement under a very low mass loss rate.
Surface temperature was measured by thermocouples during the current tests. The measured
values, for example the ignition temperature of the timber materials, are similar to the
measurements obtained by others. Since surface temperature, especially up to the moment
of ignition, plays a critical role in pyrolysis and ignition processes, current measurements
are useful to validate later computer modelling results.
148
Ej'D G.iEj'D G.iEj'D G.iEj'D G.iEM%LMEM%LMEM%LMEM%LM $DkM hQ $DkM hQ $DkM hQ $DkM hQ
Through the research discussed in the previous chapters, the following results have been
achieved. The TGA test method and the computational scheme for the kinetics have been
improved, and kinetics from traditional scheme and the “ effective” kinetics for the studied
materials have been computed. The “ effective” kinetics are believed to be more suitable to
model the decomposition in real fires. The ignition and combustion tests in the cone
calorimeter have been conducted and major parameters for describing the ignition and
combustion processes have been measured and several phenomena related to the ignition
have been identified. Enough information is now available to carry the modelling of
pyrolysis and ignition processes on to the physical platform. The simulation work will be
presented in this chapter. The simulation results, including the major critical parameters at
ignitions determined by the cone calorimeter tests, will be used to compare with the
experimental data. This comparison will be helpful for validation of various commonly
adopted ignition criteria and development a possible gas phase criterion in next chapter.
This chapter includes 5 sections. In the first section, a computer model, the Fire Dynamics
Simulator (FDS), which employs Atreya’ s one-dimensional heat transfer equations to model
the pyrolysis process, is introduced. Its principle and major features are presented and
discussed. Then the calculation process of pyrolysis and thermal parameters for the timber
materials are given in Section 6.2. The simulation results are shown in Section 6.3. The
mixture of pyrolysis products and oxidants, and distribution above the solid surface are
simulated and analysed in Section 6.4. Finally, a brief summary of the simulation results is
given in the last section of this chapter.
6.1 FDS Modelling for Pyrolysis and Combustion
As discussed in Chapter 3, Atreya’ s one-step mathematical pyrolysis model has been
adopted for the current study. Based on one set of selection criteria identified by the author,
FDS, a CFD computer model which employs Atreya’ s heat transfer pyrolysis model
(McGrattan 2004), was selected as a numerical computational tool. FDS has the major
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
149
desired features required by the research, as identified in Section 3.1.3, and is able to
generate both solid and gas phase results for the study of ignition criteria. Aimed at solving
practical fire problems in fire protection engineering, it provides the capability to study
fundamental pyrolysis and fire dynamics. This model has been adopted by a number of fire
researchers and engineers on fire problems of various scales (Chow and Yin 2002; Hamins
et al. 2004; Hietaniemi et al. 2004).
6.1.1 Principle of FDS model
Not long ago it was believed by many experts that computer models were not particularly
helpful for real fire scenario simulation due to their limited computational capability (Cox
1994). As a result of the continuous upgrading of the computational power of modern
computers and development in modelling methodology, today’ s fire models have reached a
new level. FDS is a leader of these models and is believed to be capable of predicting fire
growth and spread (Babrauskas 2002).
FDS was developed by the Fire Research Division, National Institute of Standards and
Technology (NIST), of the US Department of Commerce and released publicly in 2000
(McGrattan et al. 2000). The current version of FDS is 4.0.5 (McGrattan and Forney 2005).
FDS is a computational fluid dynamics (CFD) model of fire-driven fluid flow. It solves a
form of the Navier-Stokes equations numerically for low-speed, thermally-driven flow with
an emphasis on smoke and heat transfer from fires. The turbulence is treated by means of
the Smagorinsky form of Large Eddy Simulation (LES) which has been briefly discussed in
Section 2.2. It is also possible to conduct a Direct Numerical Simulation (DNS) if a fine
enough numerical grid is adopted. A mixture fraction combustion model is used in FDS to
calculate combustion rate and all of the major products. It solves radiative heat transfer by
the equation using a Finite Volume Method (FVM) technique. The simulation results are
visually presented by the SmokeView software which was also developed by the same
authors (Forney and McGrattan 2004).
Details of the FDS model can be found in its manual and technical guide. Some major
features will be further discussed later in this chapter. A brief overview of its history and
major features are presented as follows.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
150
6.1.1.1 History of major developments
In FDS v1, radiative fluxes were computed with a Monte Carlo style ray-tracing from the
burning particles to the walls. This model neglected gas-to-gas and wall-to-wall interactions,
and performed poorly for a fire scenario with very hot gas layers or surfaces (Floyd et al.
2003).
In FDS v2, a new combustion model and a new radiation model were added. The new
combustion model was a mixture fraction model modified to work on the coarse grids which
were often used in FDS simulation. It built a tracking ability for major species, and also for
minor species once appropriate state relationships could be developed. The new radiation
model was based on the Finite Volume Method (Raithby and Chui 1990). It allowed FDS to
model gas-to-gas and gas-to-surface radiation heat transfer.
In FDS v2.2, a sub-combustion model was introduced temporarily to control modelling of
the pyrolysis process by modifying the kinetics for the studied materials. This function
resulted from cooperating research between the NIST and VTT (Technical Research Centre
of Finland) (Hostikka 2002).
A multi-blocking grid system was introduced in FDS v3.0 to enable different grid
resolutions within a computational domain. This feature is useful in handling cases where
the computational domain is not easily embedded within a single mesh. It is also useful for
achieving fine grid resolution for interesting sub-domains while obtaining acceptable
computational speed by setting other sub-domains with coarse grid resolutions.
A beta version of v4.0 was first released on January 2004 and the formal version was made
available in August 2004. The new features in FDS v4 related to the pyrolysis process are as
follows.
A char model has been implemented in which a thin pyrolysis front is tracked
inside a solid fuel. This front separates virgin fuel from the changed portion. The
details of the pyrolysis model with charring will be described later in this section.
The pyrolysis kinetics, activation energy E and pre-exponential factor A, can be
calculated by a default method or input by users for various types of fuel, which
will be described later in detail.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
151
Temperature-dependent material properties can be prescribed for solids in input
files. This function can be applied in the char model to set properties of both
virgin and charred fuels as a function of temperature.
6.1.1.2 Mixture fraction combustion model
The foundation of the mixture fraction combustion model is based on the following key
assumption (McGrattan 2004):
The large-scale convective and radiative transport phenomena can be simulated
directly, but physical processes occurring at small length and time scales must be
represented in an approximate manner.
The actual chemical dynamics that control the combustion energy release in fires are not
fully understood. Even if they were known, the spatial and temporal resolution limits
imposed by both present and foreseeable computer resources make it very difficult to get a
detailed description of the combustion. Thus, in FDS, the combustion is assumed to be
mixture-controlled. This implies that all the production of species of interest can be
described in terms of a single mixture fraction Z. The mixture fraction is a conserved
quantity representing the fraction of material at a given point that originated as fuel.
The relations between the mass fraction of each species and the mixture fraction are known
as “ state relations” in FDS 4 (Appendix C.7) (McGrattan and Forney 2005). These state
relations can be calculated from the chemical composition of the fuel. The oxygen mass
fraction in the state relation provides the information needed to calculate the local oxygen
mass consumption rate. The form of the state relation that emerges from classical laminar
diffusion flame theory is a piecewise linear function. This linear function will be used to
determine the full history in a state relation chart, such as in Figure 6.1. Then the local heat
release rate is computed from the local oxygen consumption rate at the flame surface by the
oxygen consumption method used in the cone calorimeter.
Generally, a combustion reaction can be expressed as follows:
ProductsvOvFuelv iPiOXYF ,2 Σ→+ (6.1)
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
152
where the values of vi are the stoichiometric coefficients for species i, including the fuel,
oxygen and products in the overall combustion process.
The mass consumption rates for fuel and oxidizer have the following relation:
oxyoxy
oxy
FF
F
MWv
m
MWvm ''''
= (6.2)
where MWF and MWoxy are the fuel and oxygen molecular weights.
The scalar parameter, mixture fraction Z, is defined as:
( )
∞
∞
+−−
=,,
,
oxyIF
oxyoxyF
YsY
YYsYZ (6.3))
where:
F
oxy
FF
oxyoxy
m
m
Mv
Mvs
′′′′
== (6.4))
s: the ratio of oxidizer and fuel consumption rates in a stoichiometric reaction;
iY : the mass fraction of species i;
IFY , : the fraction of fuel in the fuel stream from a fuel/pyrolysis source;
oxyY : the mass fraction of oxygen, and
∞,oxyY : the ambient oxygen mass fraction.
Therefore, the mixture fraction value will be 1 in a region containing only fuel, and 0 where
the oxygen mass fraction takes its undepleted ambient value, ∞,oxyY . Therefore the fuel mass
fraction is zero, and will vary between 0 and 1 between such regions.
It is assumed in the mixture fraction model that the chemical reaction between the
consumed fuel and oxidizer occurs so rapidly that the fuel and oxidizer cannot co-exist. This
means that the fuel and oxidizer simultaneously vanish at a flame surface:
∞
∞
+==
,,
,;oxyIF
oxyff YsY
YZZZ (6.5)
where Z is the mixture fraction, a distribution function in space and time.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
153
For a normal oxygen containing hydrocarbon fuel, CxHyOz, its ideal reaction in the normal
atmospheric environment is expressed as follows:
( )
222
22
76.3242
76.324
Nzy
xOHy
xCO
NOzy
xOHC zyx
×
−+++
→+
−++ (6.6)
The ideal mixture fraction for the fuel is the “ ideal” fuel/air ratio for the reaction, and this
calculation is given as:
( )22
,
76.324
NOzy
xOHC
OHC
massmassmass
massmass
Z
zyx
zyx
airF
F
total
FidealF
+
−++=
+==
(6.7)
where:
OHCF MWzMWyMWxmass ⋅+⋅+⋅= ;
( )NOair MWMWzy
xmass 276.3224
×+
−+= ;
MWC, MWH, MWO, and MWN are molecular weights for elements C, H, O and N
respectively.
This ideal mixture fraction will be used to determine the turning point in the state relation
chart and therefore further determine all the combustion products as functions of the mixture
fraction Z.
6.1.1.3 Pyrolysis and charring models in FDS
The heat transfer and pyrolysis in a charring fuel are described by a one-dimensional model
which was originally developed by Atreya and described in Section 3.2. To perform a
numerical solution, further simplification was made. The pyrolysis is assumed to take place
over an infinitesimally thin front. The model includes the conduction of heat inside the solid,
the evaporation of moisture and decomposition of the virgin material to gaseous fuel and
char. The volatile gases are instantaneously transported to the surface.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
154
Based on the improvement made by Ritchie et al. (1997), the basic governing equations
including consideration of moisture release were shown as Equations (3.6) to (3.9). The
pyrolysis rate of the material was modeled with a single step first order Arrhenius reaction:
RTEcIsolid eAm /
, )( −−=′′ ρρ (6.8)
where:
A: pre-exponential factor, m/s;
Isolid ,ρ : initial solid density, kg/m3;
cρ : density of char, kg/m3;
E: activation energy, kJ/mol⋅K;
Ts: surface temperature, K.
The A and E are chosen so that the pyrolysis takes place very close to the desired pyrolysis
temperature for the solid fuel. The thin pyrolysed front is moving inside the material. The
velocity of the front, v, is given by
( )cIsolid
mv
ρρ −′′
=,
(6.9)
6.1.1.4 Ignition criteria in FDS
The ignition criteria are not directly documented in FDS. One convenient method to “ judge”
the ignition is via visual observation by the software SmokeView. The parameters normally
provided in FDS to form the “ isosurface” are heat release rate per unit of volume (HRRPUV)
and mixture fraction. There should be some internal criteria (not discovered and controlled
by users at the moment) for these parameters to present the flame. While FDS does not give
criteria of ignition directly and leaves great flexibility for users to pick up some suitable
parameters for the prediction purpose, discussion of ignition phenomena will start from the
available parameters in FDS: the HRRPUV and mixture fraction.
When these two parameters were used for the visual observation, there are some limitations.
First, the observation of flame is just an “ apparent” method. When different values of
parameters were set up, the shape as well as the moment of flame appearing will vary
largely. Secondly, the observations by different parameters for a simulation may be different.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
155
Therefore, in current thesis, the ignition determined by visual observation in FDS will be
treated as “ apparent ignition” and only used for comparison purpose. The detailed
discussion of limitations of a critical HRR as the criterion of ignition will be further
performed in Section 7.1. All the ignition parameters in simulation will be extracted at the
moment of ignition decided by previous corresponding cone calorimeter tests.
The combustion model adopted in FDS is a mixture fraction model. When Atreya’ s
pyrolysis model was used in FDS, the pyrolysis rate of solid fuel can be obtained through
modelling. The mixture fraction is able calculated from the pyrolysis rate by equations in
Section 6.1.1.2. Therefore current author assumes that the criteria of ignition in FDS are
mainly related to the mixture fraction. This assumption of value in the simulation domain
has also been verified by personal correspondence (McGrattan 2004). While a critical
mixture fraction is a materials property under certain pyrolysis conditions, this supposed
criterion needs to be further studied and the prediction of ignition in FDS can be improved.
To find a more suitable criterion for the studied materials, particularly the mixture fraction,
is one of the major aims of the current research. Therefore, in the later stage of this thesis,
all possible criteria of ignition in FDS will be investigated in detail. These criteria include
critical surface temperature, critical mass flux, critical heat release rate, and the mixture
fraction.
6.1.2 Treatment of thermal parameters for timber in FDS simulation
Calculations of the mixture fraction and thermal dynamic parameters for the current studied
timber materials are described in the following sections.
6.1.2.1 Mixture fraction relation for timber
Even though the compounds of timber material are very complex, as indicated in Section
2.3.3.1, certain simplifications must be made for modelling study. A typical expression for
the chemical compound for timber material is C3.4H6.2O2.5 (Ritchie et al. 1997) which was
adopted in the FDS model. Applying Equation (6.6) to this expression, the reaction is given
as follows:
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
156
( )
222
225.22.64.3
76.37.31.34.3
76.37.3
NOHCO
NOOHC
×++→++
(6.10)
The ratio of consumption rates for the fuel and oxygen, s, can be calculated from Equation
(6.4).
36.187
327.3 =×==FF
oxyoxy
Mv
Mvs
Substituting the s value and 233.0, =∞oxyY into Equation (6.3) yields the expression of
mixture fraction for timber:
233.036.1
233.036.1
, +−+
=IF
oxyF
Y
YYZ (6.11)
The ideal mixture fraction for timber, Zwood,ideal which is the stoichiometric ratio of fuel to
air, is calculated using Equation (6.7).
12.0)2876.332)(
25.2
42.6
4.3(87
87
,
=×+−++
=
+=
airwood
woodidealwood massmass
massZ
The state relations for wood were then determined from Equation (6.10) and shown in
Figure 6.1. It is noted that at Z=0 (no fuel present), the mass fractions of all gases are
obtained from composition of the atmosphere, and at Z=1, only fuel is present. At the ideal
mixture fraction, Z=0.12, the fuel and oxygen have been used up, and only combustion
products are present. At the two regions between these three points (Z=0, 0.12, and 1) linear
relationships exist for all reactants and products, as indicated previously.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
157
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Z
Mas
s Fr
actio
n
O2
FuelN2
CO2
H2O
Figure 6.1 State relations for wood
6.1.2.2 Calculation for the thermal dynamic parameters
FDS adopts a pyrolysis model as presented in Section 4.2.2.2. Solid fuels are treated as
either thermoplastics or charring materials. Thermoplastic materials are assumed to pyrolyse
at the surface and charring materials at a pyrolysis front that propagates through the solid.
The burning rate (pyrolysis rate) for these two types of solids is given by Arrhenius’
equation as follows:
For the thermoplastic RTEseAm /−=′′ ρ (6.12)
For the charring material RTEcs eAm /)( −−=′′ ρρ (6.13)
where:
A: pre-exponential factor, m/s;
E: activation energy, kJ/mol;
ρs: the surface density, g/m3;
ρc: the char density, g/m3;
Ts: the surface temperature, K;
R: universal gas constant, 8.314×10-3 kJ/mol⋅K;
m ′′ : mass flux, g/m2⋅s.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
158
There are two methods to calculate these parameters in FDS. They are described below,
using charring materials as an example.
The default method in FDS is to use a set of given values, 2.6E8 kg/m2⋅s for product of the
pre-exponential factor A and solid density, and 20 g/m2⋅s for the critical mass flux (Hostikka
2004). The ignition temperature for the specific material is provided by the user via the
database file. Then the activation energy is calculated through the above formula (Equation
(6.13)) at ignition point.
( )
−′′
=cs
crig A
mRTE
ρρln (6.14)
With heat transfer calculation, this computed activation energy is then used to calculate the
pyrolysis rate as well as the mixture fraction. When the ideal mixture fraction value is
reached in the simulation domain, ignition takes place immediately.
An alternative method is to supply values of the E and A as inputs. The values can be
obtained from other sources, such as the TGA tests. Then from Equation (6.13), mass flux is
calculated by the kinetics as a function of the surface temperature. Ignition takes place when
the ideal mixture fraction is reached at some point in the computational domain.
The first method relies on the input of three parameters, A, crm ′′ and Tig, which are difficult
to obtain from experiments. FDS applies a fixed pre-exponential factor to the whole range
of materials, which is questionable since A is strongly material-dependent as demonstrated
in Table 4.3 of this thesis and in the work of Moghtaderi (2001).
The difficulty for the second method is the determination of E and A. Although these
parameters can be obtained from the TG analyser, the application of these basic thermal
kinetic parameters in real fire modelling is still very limited. Generally, as discussed in
Section 4.2.3, the difference between fundamental thermal dynamic testing conditions and
real fire scenarios, both in bench and full scales, is the first concern in applying these
kinetics (i.e. E and A). Secondly, large variations in the values of these parameters have
been reported from different researchers using different testing equipments and test methods
even for a single material (Hostikka 2002). Thirdly, lack of validation of those parameters in
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
159
complicated fire scenarios, mainly through lack of suitable calculation tools/computer
models, impedes their application (Hostikka 2002).
It is necessary to note that the values of the basic thermal parameters, Tig and crm ′′ , adopted
in FDS, are suggested for general evaluation purposes only. They may not be applicable to a
specific material and fire scenario without validation (Hostikka 2004).
Simulation results obtained by adopting the above two methods are compared in the
following section.
6.2 Setting Up and Running of FDS Simulations
General information of the cone calorimeter tests that their settings were followed up in
FDS simulations were described in Chapter 5. Represented here are configurations of the
simulation domain.
6.2.1 General setting
6.2.1.1 Simulation domain and grid size
A computational domain 400 mm wide by 400 mm long by 500 mm high was used for FDS
simulations of the cone calorimeter combustion region (Figure 6.2). An air flow upward
through the domain at a flow rate of 24 l/s was specified using a vent at the top of the
domain. The four vertical glass walls were adopted as a boundary condition following the
physical configuration of the cone calorimeter. An open boundary condition was specified at
the bottom to allow a free flow of air into the domain.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
160
Figure 6.2 Simulation domain for the cone calorimeter situation
The multi-blocking technique was adopted. A grid size of 3×3×2 mm was applied to a
volume including the sample and the heater (150×150×100 mm) and the remaining part of
the domain has a resolution of 5×5×5 mm. The selection process for the grid resolution is
discussed in Section 6.2.3. An illustration showing the grid is given in Figure 6.3.
Figure 6.3 Grids for the simulation domain
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
161
6.2.1.2 “Cone” heater
The cone heater was represented by a stair polygon shape heater due to the FDS limitation
that all objects in a domain must be rectangular. The height of each level of steps is 6 mm.
Figure 6.4 shows an illustration of the heater’ s structure. The effect of the stair polygon
shape heater on the radiation heat transfer is discussed in following section.
Figure 6.4 Bottom view of the stair shape heater
6.2.2 Heater temperature and heat flux
The input temperature value for the heater in the simulation was determined by trial and
error. A heater temperature value was selected when it produced the desired heat flux at the
sample surface.
The radiative heat flux at the surface of the sample was obtained by placing heat flux gauges
at the desired locations on the sample surface. To investigate the uniformity of surface
radiation, surface temperatures at various locations on the surface were also measured in the
simulation. The radiation and temperature distribution at different distances along an axial
symmetry in the domain are shown in Figure 6.5 and Figure 6.6. In these two figures, the
rapid rises for radiation and temperature curves at about 22 seconds were caused by ignition
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
162
of the simulated timber sample when more radiative heat was fed back from the flame to the
surface of the sample.
0
20
40
60
80
100
0 5 10 15 20 25 30Time (s)
Rad
iatio
n (k
W/m
2 )
Centre-0mm
12mm
24mm
36mm
Figure 6.5 Axial distribution of radiation level
0
100
200
300
400
500
0 5 10 15 20 25 30
Time (s)
Tem
pera
ture
(oC
)
Centre-0mm
12mm
24mm
36mm
Figure 6.6 Axial distribution of surface temperature
However, the symmetry of the radiation and temperature of the sample surface was only
approximate. Wilson et al. (2003) studied surface radiation distribution in the cone
calorimeter. Their theoretical calculation and measurement by radiative flux gauges are
shown in Figure 6.7. The cone heater surface temperature was set to 650 oC. The radiative
flux on the central part of the sampling area, which was 25 mm below the bottom of the
heater, was measured as 32.7 kW/m2. The unit for the contour curves shown in the figures
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
163
was W/m2. Followed the experimental set up, the surface radiation under same heater
temperature was monitored in a FDS simulation, shown in Figure 6.8.
(a) Theoretical (b) Measured
Figure 6.7 Theoretical and measured irradiance distributions for the cone calorimeter (source: Wilson et al. (2003))
Figure 6.8 Boundary output for surface radiation
It is found that the boundary radiative flux contour on the surface of the sample did not
show perfectly circular shapes. Compared between these figures, it is obvious that the
radiative heat flux at the edge areas is lower than the experimental value. While the
pyrolysis process in the central area is the major concern of ignition study and the heater
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
164
temperature was calibrated by the central radiation, lower values obtained for these global
parameters, such as HRR, may occur in the simulation.
6.2.3 Cell grid
Grid dependence is an issue that needs to be addressed when using numerical simulation
methods (Cox and Kumar 2002). A number of studies on this issue have been done by
researchers for various fire source and geometry conditions (Friday and Mowrer 2001;
Heskestad 2002; Ierardi and Barnett 2003).
In the current research, grid sensitivity studies were performed. The grid was systematically
refined until the output quantities did not change appreciably with each refinement. A set of
preliminary simulations was carried out to decide the appropriate grid settings for the FDS
model. The simulations were divided into two steps. In the first step, only change of grid on
direction of axis Z was applied. Settings of this part preliminary simulation are shown in
Table 6.1. After that, horizontal grid was investigated in the second step by fixing grid on Z
axis at a value determined in the step one. Table 6.2 shows settings for the second step
simulations. It is noticed that the height of the 2nd sampling point above the sample surface
could not be exactly identical under various vertical grid resolutions. Therefore, differences
of those values were reduced to minimum by choosing a proper height.
Table 6.1 Settings for the first step preliminary simulations
Vertical (z axis)
Other axies
RES-V1 1 3 30 Surface 15RES-V2 2 3 30 Surface 16RES-V3 3 3 30 Surface 15RES-V5 5 3 30 Surface 15
Simulation code
1st sampling point
Grid resolution (mm) Distance between heater and sample
(mm)
Height of 2nd sampling point
(mm)
Table 6.2 Settings for the second step preliminary simulations
Horizontal Axes
Vertical (z axis)
RES-H1 1 2 30 SurfaceRES-H2 2 2 30 SurfaceRES-H3 3 2 30 SurfaceRES-H4 4 2 30 Surface
Simulation code
Grid resolution (mm) Distance between heater and sample
(mm)Sampling point
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
165
The sample material is pine wood and the default material property inputs given by the FDS
in the database were adopted. Parameters were monitored at two sampling points (one on
the sample surface and another in the space above the sample) to check both solid and gas
phase heating effects. The parameters monitored include temperature, radiative flux, mass
loss flux, and fuel concentration (expressed as the mixture fraction (Z)). These parameters
were sampled at second intervals until ignition occurred. Table 6.3 shows some results from
the first step preliminary simulations.
As shown in Figure 6.9 and Figure 6.10, histories of the simulated surface temperature and
mass flux in the cone calorimeter situation were obtained in different cell resolution settings.
The parameters were obtained at the point 1, centre of a pine wood sample surface.
Table 6.3 Major results from the first step preliminary simulations
T-1 Radiation-1 Mass flux-1 T-2 Z-2
(oC) (kW/m2) (kg/m2s) (oC) (kg/kg)
1 207.5 67.6 2.9E-06 119.0 1.4E-05
2 210.2 58.8 3.5E-06 118.9 1.7E-05
3 243.5 57.1 4.4E-06 111.8 2.2E-05
4 215.2 56.3 5.8E-06 89.5 4.2E-055 197.7 51.8 3.2E-06 25.5 2.6E-06
T-1 Radiation-1 Mass flux-1 T-2 Z-2
(oC) (kW/m2) (kg/m2s) (oC) (kg/kg)1 299.4 84.1 5.0E-04 164.8 2.0E-032 303.3 70.0 5.9E-04 184.4 2.2E-033 272.1 68.9 8.2E-04 200.6 3.0E-034 318.6 71.9 1.3E-03 233.8 5.8E-035 296.5 56.5 4.1E-04 157.1 2.6E-03
tig Tig,-1 Mass flux-1 T-2 Z-2
(s) (oC) (kg/m2s) (oC) (kg/kg)1 11.0 320.0 1.3E-03 200.4 4.5E-032 10.9 320.7 1.3E-03 195.5 3.9E-033 10.5 290.2 1.3E-03 210.3 3.9E-034 10.0 318.6 1.3E-03 233.8 5.8E-035 11.8 321.7 1.3E-03 169.5 4.6E-03
At 10s
At Ignition
Grid (mm)
At 5s
Note: T, t and Z are temperature, time and mixture fraction respectively;
-1, at point 1 (on the sample surface); -2, at point 2 (above the sample surface).
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
166
0
100
200
300
400
500
0 2 4 6 8 10 12 14 16
Time (s)
Tem
pera
ture
(oC
)
V-1mm
V-2mm
V-3mm
V-5mm
Figure 6.9 Surface temperature from various vertical grid resolutions (pine, 50 kW/m2)
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0 2 4 6 8 10 12 14 16
Time (s)
Mas
s flu
x (k
g/m
2 s)
V-1mm
V-2mm
V-3mm
V-5mm
Figure 6.10 Mass flux from various vertical grid resolutions (pine, 50 kW/m2)
From the above tables and figures it can be seen that differences for all the simulated
parameters between 1 mm and 3 mm vertical grid resolutions are relatively small, especially
between those with 1 mm and 2 mm resolutions, except for the time after 11 seconds – the
moment of the “ apparent ignition” occurred. Two key parameters for the pyrolysis process
from those two resolutions, surface temperature and mass flux, match quite well with each
other during the pyrolysis process. In regard to the gas phase, the values for the mixture
fraction, which is a critical parameter in FDS to describe pyrolysis, were also very close to
each other for the 1 mm and 2 mm grids, at 10 seconds and the moment of the “ apparent
ignition” as shown in Table 6.3. Considering the simulation time for 1 mm grid is about 16
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
167
times longer than that for 2 mm resolution, the 2 mm grid resolution was chosen in all
simulations to achieve a good balance between satisfactory simulation accuracy and
acceptable simulation time.
Major comparison results of the second step simulation, varying grid resolutions
horizontally, are shown in Figure 6.11 and Figure 6.12. The differences of monitored
surface temperature and mass flux between 2 and 3 mm grid resolutions are relatively small.
Similar consideration exists for the 1 mm horizontal grid resolution as that for the vertical
values. Therefore, 3 mm horizontal grid resolution was chosen in all the late simulations.
0
50
100
150
200
250
300
350
400
450
0 2 4 6 8 10 12 14
Time (s)
Tem
pera
ture
(o C
)
H-1mmH-2mmH-3mmH-4mm
Figure 6.11 Surface temperature from various horizontal grid resolutions (pine, 50 kW/m2)
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0 2 4 6 8 10 12 14
Time (s)
Mas
s flu
x (k
g/m
2 s)
H-1mmH-2mm
H-3mmH-4mm
Figure 6.12 Mass flux from various horizontal grid resolutions (pine, 50 kW/m2)
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
168
6.2.4 Summary of input parameters in FDS simulation
A summary of major inputs and settings in the simulations are shown in Table 6.4. As
measured in the cone calorimeter tests (see Section 5.2.2), the moisture content of 8.5% was
adopted for the mountain ash in these simulations. Densities of the wood and char were
adopted from the measurements made in the cone calorimeter tests. The heater temperature
was set to achieve the desired radiation levels in the table by the method introduced in
Section 6.2.2. The default values for E and A were generated by the model according to the
first method described in Section 6.1.2.2. Other E and A values for mountain ash were
obtained from the TG tests, both lower and higher heating rates, shown in Table 4.3 and
Table 4.4. For the pine, the obtained kinetics of from the lower heating rate range are close
to the values from Atreya’ s study (Atreya 1984). Therefore values of E and A suggested by
Atreya, 5.1E+11 1/s and 126 kJ/mol for the A and E respectively, were used for the pine in
the current simulations. Those values were also adopted in FDS as default values (Hostikka
2004). The ignition temperature was the default setting in FDS. Two thicknesses of the
sample were simulated. Besides the parameters listed here, values for all the other
parameters, such as heat of combustion, heat conductivity, were adopted from the FDS
database for the woods.
Table 6.4 Major inputs and settings for simulations
Simulation code
Minimum Grid (mm)
MaterialExternal
Radiation (kW/m2)
E (kJ/mol)
A (1/s)
Tig
(oC)Thickness
(mm)
SIMU-01 140.2 1.2x108 32
SIMU-02 177.4 6.7x1011 32SIMU-03 140.2 1.2x108 32SIMU-04 177.4 6.7x1011 32
SIMU-05 140.2 1.2x108 14
SIMU-06 Default Default 390 32SIMU-07 Default Default 350 14SIMU-08 140.2 1.2x108 32SIMU-09 177.4 6.7x1011 32SIMU-10 30 126 5.1x1011 32SIMU-11 50 126 5.1x1011 32
3x3x2
Pine
75
30
50mountain
Ash
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
169
6.3 Simulation Results
Following the configuration described in Section 6.2 (including Table 6.4), a series of
simulations were carried out to simulate the pyrolysis process for timber. Major simulation
results are shown as below and some comparisons are given between the simulations and
the cone calorimeter test results. The kinetics adopted were the “ effective” values obtained
from higher heating rate range, i.e. 140.2 kJ/mol and 1.2x108 1/s. Simulation results from
these kinetics were illustrated in following sub-sections unless specified.
6.3.1 Observation of initial flame development
Although in a cone calorimeter test it is possible that combustion can start anywhere on a
timber surface due to non-uniform properties of the specimen, it was observed that the first
flame always appeared near the centre of the sample surface in the FDS simulations. The
initial development of the flame on the sample surface is shown in Figure 6.13 for the
mountain ash under 30 kW/m2. The parameter used to represent the flame is the heat release
rate per unit volume, HRRPU. The time interval between these pictures is 0.5 s.
Sets of photos taken from the piloted cone calorimeter tests using the mountain ash are
shown in Figure 6.14 for 30 kW/m2 radiation level and in Figure 6.15 for 50 kW/m2
radiation level. The camera speed used was 15 frames per second (approximate as an
interval of 0.06 s).
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
170
(a) (at tig) (b) (at tig+0.5s)
(c) (at tig+1s) (d) (at tig+1.5s)
(e) (at tig+2s) (f) (at tig+2.5s)
Figure 6.13 Simulated flame development for the mountain ash under 30 kW/m2
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
171
(a) (at tig-0.06s) (b) (at tig)
(c) (at tig+0.06s) (d) (at tig+0.12s)
(e) (at tig+0.18s) (f) (at tig+0.24s)
(g) (at tig+0.3s) (h) (at tig+0.36s)
Figure 6.14 Flame development in test from the mountain ash under 30 kW/m2
Notes for Figure 6.14:
(a): before ignition, pyrolysis products was observed within the gas stream
(b): initial flame appeared on the surface
(c) ~ (e): flame spread on the surface
(f) ~ (h): flame fully developed
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
172
(a) (at tig-0.12s) (b) (at tig-0.06s)
(c) (at tig) (d) (at tig+0.06s)
(e) (at tig+0.12s) (f) (at tig+0.18s)
Figure 6.15 Flame development in test from the mountain ash under 50 kW/m2
Notes for Figure 6.15:
(a) ~ (b): moments before ignition, pyrolysis products appeared within the flow
stream with a higher concentration level compared to the 30 kW/m2 in test
(c): initial gas phase flash (colored blue and as indicated by an arrow), the first
flash occurred above the cone heater in the gas phase
(d) ~ (f): flame fully developed without obvious surface spreading
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
173
From above illustrations it can be seen that the general trends of flame developing in the
modelling match roughly with the real observation. However a perfect matching in flame
shape and development is either difficult or unnecessary. The simulated flame shape
provides only a visual observation of contour of heat release rate since the chosen value to
display the flame will affect flame shape greatly. Meanwhile, there are still some
discrepancies and limitations in FDS simulation. First, the initiation of combustion in
simulation always occurs in the centre of upper sample surface. In reality, the first flash can
appear on any point of upper sample surface, even in the gas phase as that discussed below.
Second less gas phase combustion (shown as the flame height in Figure 6.13 (e) and (f)) in
simulation was observed compared to that in test (shown as in Figure 6.14 (g) and (h)). The
reason to cause this difference is the mixture controlling mechanism (infinite fast mixture
speed) in FDS. Thirdly, the HRRPUV used to form the flame contour is also affect the
flame appearing and its shape greatly, as discussed in the last sub-section.
One interesting phenomenon is that the initial flash appeared just above the heater in the test
under 50 kW/m2 radiation (Figure 6.15(c)). It is speculated that this was caused by different
conditions at top of heater and spark plug. As discussed in Section 2.4.3, key factors for an
ignition occur include fuel, oxygen, heat and atmosphere. Therefore for a piloted ignition in
cone calorimeter environment can be simplified as suitable fuel concentration and enough
temperature at the spark plug location. This suitable fuel concentration should be between
the lower and upper flammable limits. In the cone calorimeter, the solid surface was heated
by the radiative flux from the heater and generated pyrolysed products. When the pyrolysis
rate increased quickly before ignition (under a high radiative flux) and a small volume of
pyrolysed products with the suitable fuel concentration rise up rapidly, the pulsing spark
may miss out this volume. This phenomenon is really possible when turbulent flow exists
above the solid surface. Then the spark plug becomes surrounded by rich fuel stream (above
the upper flammable limit). Above the top of the heater, the pyrolysis stream has more
chance to mix with fresh air and fell into a suitable range for combustion again. The gas has
been also heated up enough when it passed through the heater. Then ignition occurred. This
observation matches with the phenomenon that there is a small peak in the mass loss rate
measurement for the mountain ash in the cone calorimeter test in Figure 5.14. It is also
accord with the theoretical explanation made in Chapter 5 for presence of the peak. As
discussed before, for the timber materials, the mass flux before the ignition, which includes
the release of moisture and inert gases, is relatively high.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
174
6.3.2 Surface temperature
The surface temperature was monitored in all simulations. Shown in Figure 6.16 to Figure
6.18 are simulated temperatures at the central point on the mountain ash sample surface at
different radiation levels, compared to that measured in the cone calorimeter tests. The
simulated critical surface temperatures, i.e. the surface temperature at the ignition times, are
shown in Table 6.5 and Table 6.6 for the mountain ash and pine respectively.
0
100
200
300
400
500
600
0 2 4 6 8 10 12 14 16
Time (s)
Tem
pera
ture
(o C
)
Tested 1Tested 2Simulated
Figure 6.16 Measured and simulated surface temperatures under 75 kW/m2
(Average tested tig=10.7 s, Tig=290 oC)
0
50
100
150
200
250
300
350
400
0 5 10 15 20 25 30 35Time (s)
Tem
per
atu
re (
oC
)
SimulatedTested 1Tested 2Tested 3
Figure 6.17 Measured and simulated surface temperatures under 50 kW/m2
(Average tested tig=27.0 s, Tig=334 oC)
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
175
0
50
100
150
200
250
300
350
400
450
0 10 20 30 40 50 60 70 80 90 100Time (s)
Tem
pera
ture
(oC
)
SimulatedTested 1Tested 2Tested 3
Figure 6.18 Measured and simulated surface temperatures under 30 kW/m2
(Average tested tig=153.2 s, Tig=330 oC)
6.3.3 Heat release rate
Figure 6.19 and Figure 6.20 show the simulated heat release rate curves of the mountain ash
under 50 kW/m2 and 75 kW/m2 in comparison with the test values. These experimental
results were obtained from the cone calorimeter tests shown in Chapter 5. The full
combustion history for the mountain ash normally lasts thousands of seconds. Due to the
limitation of computational power, it is nearly impossible to simulate the whole history of
combustion by using current grid settings. Therefore a coarser grid resolution, 3×3×4 mm,
was adopted for the long simulation under 75 kW/m2. The simulated combustion periods
were 150 and 2200 seconds for the radiative conditions of 50 kW/m2 and 75 kW/m2
respectively.
It can be seen that simulated curves have similar onset times of ignition and trends to the
experimental results, except for the second peak in the HRR curves. The times for the first
peak in both simulations and tests agree with each other. However by observing the HRR
curve in the long simulation, the second peak did not appear. This may be explained by the
fact that the FDS model does not have the capability to simulate certain changes in real
timber burning, such as bend and crack. As discussed in Section 2.3.3.2, these changes
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
176
normally increase the external surface area and expose inner unburnt virgin parts to the
radiative flux that all contribute to a higher burning rate and form the second peak.
0
50
100
150
200
250
0 500 1000 1500 2000 2500 3000
Time (s)
HR
R (
kW/m
2 )SimulatedTest1Test2Test3
Figure 6.19 HRR curves under 50 kW/m2 for the mountain ash
0
50
100
150
200
250
300
350
400
0 500 1000 1500 2000 2500Time (s)
HR
R (k
W/m
2 )
Test-1Test-2
Test-3Simulated
Figure 6.20 HRR curves under 75 kW/m2 for the mountain ash
It is clear that the simulated peak and average values of the HRR are lower than the test
results. There are a number of reasons contributing to this difference. The first is the
different heating conditions in the simulation and in the test. As mentioned in Section 6.2.2,
the radiation around the outer area is lower than that of the central area of the sample
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
177
surface. While the temperature of the stair-step shape heater was determined to ensure a
specific radiative flux was received at the central point of the surface, the total radiative heat
flux applied to the sample in the simulation is lower than that in the real test under the
“ same” radiative flux settings. This can be observed from Figure 6.7 and Figure 6.8.
Secondly, it results from the limitation of FDS to simulate the charring process in detail.
The charring sub-model in FDS is not sophisticated enough to simulate complicated details
in the charring process, such as contraction, surface recession, moisture vaporization and
cracking, etc. As mentioned in the literature review (Section 2.3.3.2), these phenomena
usually result in an enhancement of exposure of inner fresh solid and heat transfer that
contribute to a higher combustion rate and a higher heat release rate (Jonsson and Pettersson
1985).
6.3.4 Mass flux
The simulated critical mass fluxes (i.e. the mass flux at the “ ignition” occurred, the ignition
times were obtained from the cone calorimeter tests) are listed in Table 6.5 and Table 6.6
for the mountain ash and pine respectively. Shown in the tables included surface
temperature at the “ ignition” .
Table 6.5 Major inputs and results for simulation for the mountain ash
External Radiation (kW/m2)
E (kJ/mol)
A (1/s)
Tig
(oC)Thickness
(mm)tig
(s)Tig
(oC)
Mass Fluxcr
(g/m2s)
SIMU-01 140.2 1.2x108 32 326 1.0
SIMU-02 177.4 6.7x1011 32 331 1.3
SIMU-03 140.2 1.2x108 32 334 1.3
SIMU-04 177.4 6.7x1011 32 342 1.5
SIMU-05 140.2 1.2x108 14 337 1.2SIMU-06 Default Default 390 32 320 1.3SIMU-07 Default Default 350 32 283 1.4SIMU-08 140.2 1.2x108 32 343 1.6
SIMU-09 177.4 6.7x1011 32 352 1.710.7
27.0
153.2
75
50
Simulation code
Settings & Tested Results
30
Note: All simulation parameters are calculated for the central point of the sample surface.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
178
Table 6.6 Major inputs and results of simulation for the pine
External Radiation (kW/m2)
Thickness (mm)
E (kJ/mol)
A (1/s)
tig
(s)Tig
(oC)
Mass Fluxcr
(g/m2s)SIMU-10 30 32 126 5.1x1011 98.3 365 1.1SIMU-11 50 32 126 5.1x1011 19.3 354 1.2
ResultsSet-up
Simulation code
Figure 6.21 to Figure 6.23 show the initial part of mass flux curves for the mountain ash
from 30 kW/m2 to 75 kW/m2 radiative fluxes. These simulated results are also compared
with those obtained from the cone calorimeter tests. Mass flux from “ dried” timber tests
were also shown for the 30 kW/m2 and 50 kW/m2 radiative fluxes.
0
2
4
6
8
10
12
14
0 50 100 150 200Time (s)
Mas
s flu
x (g
/m2 s)
SimulatedTest-1_moistured
Test-2_mosituredTest-3_driedTest-4_dried
Figure 6.21 Simulated mass flux for the mountain ash under 30 kW/m2
compared with test data
0
2
4
6
8
10
12
14
16
18
20
0 20 40 60 80 100Time (s)
Mas
s flu
x (g
/m2 s)
Simulated
Test-1_moistured
Test-2_moistured
Test-3_dried
Test-4_dried
Figure 6.22 Simulated mass flux for the mountain ash under 50 kW/m2
compared with test data
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
179
0
5
10
15
20
25
0 10 20 30 40 50Time (s)
Mas
s flu
x (g
/m2 s)
Simulated
Test-1
Test-2
Test-3
Figure 6.23 Simulated mass flux for the mountain ash under 75 kW/m2
compared with test data
It is noticed that the simulated mass flux curves have a sharp rise starting from about 1.5
g/m2⋅s rising to 15 g/m2⋅s within approximately 2 seconds. Such rapid rises occurred in all
simulations. This rise is resulted from increased radiative flux feedback from the sustained
combustion followed the ignition. The onsets of the rising under different radiative fluxes
have similar values, ranging from 1.0 to 1.6 g/m2⋅s. Other researchers (Deepak and Drysdale
(1983), Drysdale and Thomson (1989), and Tewarson et al. (1999)), have reported a range
from 0.8 to 6.0 g/m2⋅s for this onset mass flux for timbers. Detailed data from others’ results
have been reviewed in Section 2.4.3.2. The critical mass flux at the “ ignition” increased
slightly as the external radiative flux increased. This trend matches with the observation of
others (Kanury 2002; Yang et al. 2003).
The relationship between the critical mass flux and air flow rate through the simulated
domain has been also investigated. Simulated critical mass fluxes, surface temperatures and
mixture fractions at the centre of surface under 50 kW/m2 with various duct air flow rates
are shown in Table 6.7.
Table 6.7 Effect of various air flow rates
Flow rate (l/s)
Tig
(oC)
Mass fluxcr
(g/m2s)Mixture
fractioncr (g/g)
96 343 1.3 0.02724 332 1.2 0.0246 326 1.1 0.022
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
180
It is noted that Tig and critical mass flux increased as the air flow rate increased. That means
that under a higher air flow rate, the cooling effect on the sample surface and diluting effect
for the gas phase fuel concentration all become stronger. Therefore required surface
temperature and mass flux are higher compared to that at lower flow rate condition.
However the values of the mixture fraction keep stable under various duct air flow rate.
The change in the critical mass flux caused by the air flow rate changing is relatively
smaller, compared with the results obtained by Zhou et al. (2002). This is because under the
current configuration of the simulation domain (modelling a cone calorimeter situation), the
air flow velocity just above the sample surface (within 15 mm including the spark plug
location) does not change much when the total air flow rate through the top duct varies from
4 to 96 l/s. This feature can be validated by the flow patterns under various vent flow rates
that are shown in Figure 6.24. However the general trend for the critical mass flux is still
slightly increased when the duct flow rate increased. This trend matches both modelling and
experimental results obtained by Zhou et al.
In the FIST tests carried out by Zhou et al. (described briefly in Section 4.2.2.2), it was
found that when air flow velocity over a sample surface increased from 1.0 to 1.7 m/s, the
critical mass flux increased from 1.5 to 3.0 g/m2⋅s for a laminar parallel flow, radiative
heated PMMA object. In the cone calorimeter situation, air flow has a quite different pattern
compared with the laminar parallel flow pattern in the FIST experiment. In the current FDS
simulations, both laminar and turbulent flows exist in the domain simultaneously. The gas
flow velocity above the sample is approximately between 0.2 to 0.5 m/s under various total
duct flow rates. This estimation can be proved from Figure 6.24 and also from
measurements and simulations by Tsai et al. (2001). Under this lower flow rate condition,
the effect from the changing gas flow velocity on the surface cooling is lower. Therefore the
changing of velocity may have less effect on the critical mass flux.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
181
4 l/s 12 l/s
24 l/s 48 l/s
Figure 6.24 Flow velocity (×10-3 m/s) under 50 kW/m2 with various duct flow rates
6.3.5 Direct Numerical Simulation
The direct numerical simulation (DNS) computational scheme was also investigated for the
mountain ash. The domain dimensions, heater temperature set-up and sample properties
remained exactly the same as those in the LES simulations. A grid resolution of 3×3×1 mm
for the volume including the heater and sample was adopted to form a finer grid resolution
which is generally required in DNS simulations.
Shown in Figure 6.25 are the surface temperature and mass flux curves obtained from a
DNS simulation under 50 kW/m2 radiation condition, compared with the LES scheme. It
can be seen the DNS generates slightly higher initial surface temperature and mass flux. At
the ignition (27.0 s) the surface temperature in LES simulation becomes higher than that in
DNS simulation. This is caused by an earlier ignition determined by internal criteria in FDS
in the LES simulation. Generally the parameters from these two schemes at ignition are
relatively close.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
182
0
50
100
150
200
250
300
350
400
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Time (s)
Sur
face
Tem
pera
ture
(o C
)
0.0
0.4
0.8
1.2
1.6
2.0
Mas
s Fl
ux (g
/m2 s)
Temperature-DNS
Temperature-LES
Mass Flux-DNS
Mass Flux-LES
Figure 6.25 DNS results for the mountain ash under 50 kW/m2
The computational time required for the DNS computation is approximate 6 times longer
than that of LES method. Therefore, the DNS method was not used for other simulations in
current study.
6.3.6 Comparison on results from different E and A values
The major simulated parameters from adopting different E and A values are compared with
that from the tests, shown in Table 6.8. Ignition times determined by the cone calorimeter
tests and FDS simulations are all listed for comparison. It is found that ignition parameters,
such as critical surface temperature, critical mass flux and ignition time, are affected greatly
by the values of the E and A.
Adopting the “ effective” kinetic parameters E and A can generate quite good simulated
results for the ignition parameters (except a little higher surface temperature at 75 kW/m2).
Applying the “ perfect” kinetics obtained from normal used lower heating rates generated
slower pyrolysis and ignition processes, and only applicable to the low external radiation
condition. On the other hand, if the default computational scheme in FDS was adopted,
shown for 75 kW/m2 as an example, the simulated critical surface temperature and mass
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
183
flux are much higher that the experimental data. This can be explained by the earlier
ignitions (earlier than that in the tests) occurred in the simulation (see from the column
shows ignition time determined by FDS). This result also approved the necessarily to adopt
kinetics from the improved TGA test method and computation scheme when detailed
pyrolysis is to be studied in FDS.
Table 6.8 Comparison of effect of different E and A values for the mountain ash Test Results
Radiation (kW/m2)
E (kJ/mol)
A (1/s)
tig
(s)Tig
(oC)
Mass Fluxcr
(g/m2s)tig
#
(s)
Tig
(oC)
140.2 1.2x108 326 1.0 110.0177.4 6.7x1011 319 0.9 131.3140.2 1.2x108 337 1.3 26.7177.4 6.7x1011 326 1.1 29.5
Default Default 356 1.8 7.8140.2 1.2x108 343 1.6 11.4177.4 6.7x1011 334 1.3 12.8
Simulation Results
29010.775
330
334
Settings
153.0
27.0
30
50
Note: tig
# is obtained from simulations (“ apparent ignition” time).
6.3.7 Effect of sample thickness and backing material
The effects of sample thickness and backing insulation materials were investigated by a set
of simple simulations. Spruce was used in the simulation and the material properties in the
database file for the Spruce in FDS were adopted. This means that the default scheme for
the E and A values was applied. Four thicknesses were checked: 28 mm (marked as
“ thick1” ), 12 mm (marked as “ thick2” ), 8 mm (marked as “ thin1” ) and 4 mm (marked as
“ thin2” ). A flat hot plate provided two levels of uniform radiative fluxes, 68 and 53 kW/m2.
The major observations are as follows.
Figure 6.26 provides temperature curves of the “ thick1” and “ thick2” samples under the 68
kW/m2 radiation. Under the same radiative flux, the surface temperatures of the various
sample thicknesses are the same, no matter if an insulation material is present or not. A
sample temperature 1 mm below the surface remains identical with and without the
insulation material. A thicker sample (“ 1mm-thick1” in the figure) has a higher inside
sample temperature at 1 mm deep when the back of the sample was exposed, since the heat
flux within the thinner sample is easily lost from the exposed back.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
184
0
50
100
150
200
250
300
350
400
0 1 2 3 4 5 6 7 8 9Time (s)
Tem
pera
ture
(oC
)
0mm-thick2(exposed)1mm-thick2(exposed)0mm-thick1(exposed)1mm-thick1(exposed)0mm-thick2(insulated)1mm-thick2(insulated)
0mm curves
1mm curves
Figure 6.26 Simulations for the spruce under 68 kW/m2
The effect of the thickness was also checked on the thinner samples, “ thin1” and “ thin2” .
The selection of these thicknesses was based on the cone calorimeter results from the WPI,
which have been discussed in Chapter 5. The thickness that may cause different behaviours
was around 5 mm from those reported data. Shown in Figure 6.27 are inside temperatures
(at 1 and 4 mm) monitored from the simulations with the insulation backing material.
It was noted that with the insulated backing material, the thinner sample has higher inside
sample temperatures. However the temperature difference at 1 mm deep for the whole
simulation duration is quite small. In fact the temperature difference at 1 mm deep is also
very small before the “ apparent ignition” in FDS occurs (at 6 seconds).
0
50
100
150
200
250
300
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Time (s)
Tem
pera
ture
(oC
)
1mm(thin1)4mm(thin1)1mm(thin2)4mm(thin2)
Figure 6.27 Simulations for the thinner spruce samples under 53 kW/m2
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
185
Shown in Figure 6.28 are the temperature and surface mass loss rate curves from
simulations for the thinnest sample (“ thin2” ) under the two radiation levels (“ HRR-H” as 68
kW/m2 and “ HRR-L” as 53 kW/m2). The “ apparent ignition” times for the two simulations
are 6 and 9 seconds respectively. The surface temperatures at the ignitions are identical
(around 300 oC). However the inside temperatures at ignition are different: lower external
radiation generates a higher inside temperature (150 oC vs 110 oC) since a longer “ ignition”
time exists. Therefore a higher mass loss rate was generated from the lower radiation level
at “ apparent ignition” .
0
50
100
150
200
250
300
350
400
450
500
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Time (s)
Tem
per
atur
e (o C
)
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0.014
0.016
0.018
0.020
Mas
s Fl
ux
(kg
/m2 s)
T-0mm(HRR-H) T-1mm(HRR-H)T-0mm(HRR-L) T-1mm(HRR-L)Mass Flux(HRR-H) Mass Flux(HRR-L)
Figure 6.28 Simulation for the spruce under various radiation levels
Heating effect on the studied timbers with various backing materials in the current cone
calorimeter environment was also investigated. Temperature profiles at different depths,
including on bottom surface as 32 mm, from a pine specimen heated under 30 kW/m2 were
shown in Figure 6.29.
It can be seen that temperature rises were observed within the specimen, but not at back of
the specimen. With tested sample thickness, there is no significant temperature difference
from different backing conditions before the ignition occurred, even for this slow heating
scenario. Therefore, for a purpose of studying pyrolysis and ignition processes by FDS
simulation, there is no significant effect of selection of the backing materials. However,
when a full history of the combustion of the timber specimens is a research interest (as in
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
186
Section 6.3.3), setting of different backing condition may be critical. For such simulations,
“ Insulation” was set to follow the cone calorimeter testing condition.
Figure 6.29 Simulated temperatures of a pine specimen under 30 kW/m2
6.4 Distribution of Pyrolysis Products
“ Current filed models are based on one form or another of the ‘mixed is burnt’ hypothesis”
(Baum 2002). For example, in the FDS, a mixture fraction variable is used to characterize
combustion, by terminating burning at a preset minimum oxygen concentration. Above that
preset concentration, the burning rate is controlled by the rate that oxygen mixes into the
flame zone (McGrattan 2004).
Meanwhile, as discussed in Section 2.4.3, the gas phase concentration is of interest for the
current ignition criteria study. Therefore the gas phase distribution of pyrolysis products,
which is represented as the mixture fraction in FDS, requires to be investigated. For this
purpose, distribution of the mixture fraction was also studied through FDS simulation.
As illustrated in Figure 6.30 the mixture fraction was monitored at different locations above
the solid surface in the cone calorimeter. Location 1 is the location of the spark plug.
0
50
100
150
200
250
300
350
0 20 40 60 80 100 120 140Time (s)
Tem
pera
ture
(oC
)1mm-ins. 10mm-ins. 32mm-ins.
1mm-exp. 10mm-exp. 32mm-exp.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
187
Figure 6.30 Locations of the mixture fraction monitored (distance in mm)
The vertical and horizontal distributions of the mixture fraction under 50 kW/m2 for the
mountain ash are shown in Figure 6.31 and Figure 6.32 respectively. It is noted that the
average ignition time is 27.0 seconds from the tests.
Closer to the surface, a higher mixture fraction was monitored in the simulation (as shown
in Figure 6.31). Since the pyrolysed stream sources from the solid surface and mixes with
air along the rising path, the highest fuel concentration (highest mixture fraction value) is
expected at the surface. The mixture fraction curves at different heights have similar trends.
They all start increasing rapidly at about 25 seconds and reach a peak value at
approximately 27 seconds (the time ignition took place). When nearly all the pyrolysed
products were consumed immediately after ignition, the mixture fraction curve drops and
rise up again which was controlled by the surface pyrolysis rate.
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0 5 10 15 20 25 30
Time (s)
MF
(kg/
kg)
6mm-H
14mm-H20mm-H
Figure 6.31 Vertical distribution of the mixture fraction at 50 kW/m2 for the mountain ash
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
188
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0.014
0.016
0.018
0 5 10 15 20 25 30
Time (s)
MF
(kg/
kg)
Centre-0mm
12mm
24mm
36mm
Figure 6.32 Horizontal distribution of the mixture fraction at 50 kW/m2 for the mountain ash
Checking the horizontal distributions in Figure 6.32, the mixture fractions, within 12 mm of
the centreline are higher than that from the edge points. This is caused by a higher central
pyrolysis rate which was resulted from higher external radiative flux, and lower dilution
with fresh air. The higher radiative flux on the central area of solid surface has been shown
in Section 6.2.2.
These trends shown in the figures also exist in simulations for all other simulated materials
and radiative conditions. From these trends, following pattern exists: the higher mixture
fraction can be monitored along the centreline and the highest value is obtained at the
surface of the sample. This result will be used in the ignition criteria study in the next
chapter.
6.5 Summary of the Simulation Results
The pyrolysis and ignition processes in the cone calorimeter tests were successfully
simulated in FDS by applying the “ effective” kinetics obtained from the improved test
method and the computational scheme of the TGA test. The most simulated parameters for
describing timber ignition, such as ignition time and surface temperature, match with the
experimental results reasonably well. It has been shown that by using appropriate material
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
189
properties obtained from the TGA experiments, FDS, adopting Atreya’ s pyrolysis model, is
able to simulate the heat transfer, pyrolysis and ignition phenomena for timber in the cone
calorimeter environment.
Among the thermal properties of the simulated timber materials, the activation energy E and
the pre-exponential factor A are the most critical ones. It was found that a set of the values
obtained from the previous TGA tests under higher heating rates, as so-called “ effective”
kinetics, generated satisfactory simulation results, particularly for the ignition time and
ignition surface temperature.
Geometric parameters also affected the simulation results. The constructed heater was able
to generate the desired radiative flux at the centre of the sample surface. This ensured the
ignition phenomena (which occurred in the central area first in the simulation) was
simulated properly.
Resolution of the grid plays an important role in the accuracy of the simulation. A grid size
of 3×3×2 mm was found to get a balance best between higher accuracy and faster simulation
speed for the cone calorimeter situation.
The simulated surface temperatures in the central area prior to ignition are slightly higher
than those measured in the experiments. The simulated temperatures fall into a range, from
280 to 350 oC, that was reported by many other researchers. As the radiation level increases
the surface temperature at ignition decreases. This trend matches partially the current
experimental results and has been reported by other researchers.
The HRR curves from the simulations have similar trends to that of the tests, except for the
absence of a second peak in the simulation. The times for the first peaks are also similar to
the tested results. FDS can not simulate the second peak because the phenomena that cause
it, such as crack and bend of the sample, have not been built into the model. The predicted
HRR levels are lower than the values from the tests. The lower HRR level is caused by
lower radiative heat flux on the edge of the sample in the simulation.
Chapter 6 Simulating Pyrolysis and Ignition by FDS for the Woods ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
190
The mass flux curves in the simulations have approximate ranges for peak values with those
tested. However the simulated mass flux values before ignition are much lower than that
from the tests, because no moisture and other inert gases were included into this mass flux
in the FDS.
FDS is able to simulate the heat transfer within the sample and the effects of the sample
thickness and insulation backing material. For example, higher inside sample temperatures
were generated in a thinner sample, compared to a thick sample. However FDS can not
simulate the impeding effort of the upper char layer on heat and pyrolysis product transfer
Values of the mixture fraction at ignition time were monitored at various locations. It was
found that the highest mixture fraction value was obtained at the centre of the sample
surface. Information for the mixture fraction will be useful for the investigation of the
ignition criteria in the next chapter.
191
¡¢£ ¤ ¡¢£ ¤ ¡¢£ ¤ ¡¢£ ¤ £¥%¡¢&£¥P ¦k§M£g¥¨M©J¥%¡¥§£¥%¡¢&£¥P ¦k§M£g¥¨M©J¥%¡¥§£¥%¡¢&£¥P ¦k§M£g¥¨M©J¥%¡¥§£¥%¡¢&£¥P ¦k§M£g¥¨M©J¥%¡¥§ª©NQ©«©NQ©«©NQ©«©NQ©««=¥i¬Ok\¬M¬¥L§M©«=¥i¬Ok\¬M¬¥L§M©«=¥i¬Ok\¬M¬¥L§M©«=¥i¬Ok\¬M¬¥L§M©
In Chapter 6, several ignition parameters that are commonly used as criteria for wood
ignition in the modelled environment were investigated. Selection of these parameters was
based on previous review and discussion for the ignition criteria as well as the test and
simulation results for the materials studied. The mixture fraction was also monitored in the
simulation. In this chapter, all the variables previously mentioned as possible ignition
criteria are further examined. The relationship between the mixture fraction and the LFL is
investigated. Through comparison and discussion of the variables, the favoured ignition
criterion is then suggested.
There are three sections in this chapter. In the first section, judgements of ignition in FDS
are critically examined. Limitations of these methods are presented and discussed. A
comparison between the simulated results and experimental data is carried out for the major
ignition parameters in Section 7.2. In Section 7.3, analysis and discussion is performed for
the mixture fraction to develop a suggested criterion for wood ignition in the cone
calorimeter situation.
7.1 Further Discussion of Ignition in FDS
As discussed in Chapter 6, two commonly adopted methods in FDS for visual determination
of ignition are HRRPUV and mixture fraction. By applying different settings of the critical
value for those parameters, the observation results may vary a lot. There are some
limitations to apply these two parameters and further discussions of them are given as
followings.
Since the ignition phenomenon aimed to predict in the current research is a sustained
combustion, as discussed in Chapter 2 and 5, the HRRPUV is not a suitable criterion for this
purpose. Shown in Figure 7.1 are HRR curves for the mountain ash under various radiative
fluxes. The ignition time determined from the cone calorimeter tests are marked as solid
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
192
stars in the graphs and an “ apparent ignition” time is also marked as a hollow star for the 30
kW/m2.
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150Time (s)
HR
R (k
W/m
2 )30 kW/m2
50 kW/m2
75 kW/m2
Radiative flux
Ignition time from the tests. Ignition time from the simulation.
Figure 7.1 HRR curves and the ignition times for mountain ash
It can be seen that the values of HRR at which ignition occurred varied at different radiative
fluxes. It is impossible to find a universal value of the HRR as the criterion of ignition, even
it’ s related to another suggested gas phase criterion, critical energy density. As discussed by
Lyon (2004), this critical HRR for ignition of solids in a dynamic environment can be
calculated from the critical energy density as follow.
Q
cH
HRRairair
removecr ′′′=
ρ
(7.1)
where:
Hremove: rate of heat removed from solid surface, 10 W/m2·K in a controlled fire
calorimeter;
air: density of air, 1 kg/m3;
cair: heat capacity of air, 1 kJ/kg·K;
Q ′′′ : critical energy density, 1900 kJ/m3 for vapors of hydrocarbons containing
oxygen, nitrogen, etc.
The critical HRRs for polymers are reported from 10 to 40 kW/m2 (Lyon and Janssens
2005), and also varied under various external conditions. Even if the critical HRR values
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
193
under different radiative fluxes are similar, when different values were chosen, the ignition
times determined will vary greatly. It is noticed that under low radiative flux (30 kW/m2)
the value of HRR at ignition (150.3±63 s which was from the cone calorimeter test) is
extremely high. This is because a huge variation exists in the experimental ignition time
(from 90 to 216 s) under the low radiative flux (see Table 5.2). The pyrolysis rate in the
simulation is much quicker than the average value in the tests and the “ ignition” in FDS
occurred much earlier (as the hollow star at around 110 s).
The mixture fraction curves obtained at the location of the spark plug are shown in Figure
7.2 for the mountain ash under various radiative fluxes. The critical values of the mixture
fraction under radiative fluxes of 50 and 75 kW/m2 are similar (around 0.02 g/g). Similar to
that in HRR graphs (Figure 7.1), an odd value of mixture fraction (above 0.1 g/g) was also
observed under 30 kW/m2, which was caused by the earlier ignition in FDS. In FDS, both
HRR and mixture fraction increased quickly after the “ apparent ignition” occurred.
Therefore, the value of the mixture fraction at the moment of ignition determined by FDS is
also presented for comparison. Shown in Table 7.1 is critical mixture fraction at different
locations at ignition. The one monitored at the centre of the solid surface, marked as Z-surface
and representing the highest value in the domain, as noted in Chapter 6. Another value was
simulated at the location of the height of the spark plug, which marked as Z-plug. It is noticed
that the values of the HRR and mixture fraction under 30 kW/m2 were extracted at the
moment ignition occurred in FDS.
0.00
0.01
0.02
0.03
0.04
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150Time (s)
Mix
ture
frac
tion,
Z (g
/g)
30 kW/m2
50 kW/m2
75 kW/m2
Radiative flux
Ignition time from the tests. Ignition time from the simulation.
Figure 7.2 Mixture fraction at spark plug and the ignition times for the mountain ash
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
194
Table 7.1 Values for the critical parameters at ignition for the mountain ash External
Radiation (kW/m2)
tig
(s)Tig
(oC)
Mass Fluxcr
(g/m2s)Z-surface
(g/g)Z-plug
(g/g)
30 110.0# 326 1.0 0.027 0.02050 26.7 334 1.3 0.024 0.01875 11.4 343 1.6 0.030 0.023
Note: #: ignition time determined from the simulation.
It can be seen that the critical mixture fractions at the two locations are relatively similar
under various radiative fluxes at the different “ ignition” times. This provides basis for
adopting the mixture fraction as an ignition criterion. However there is still a limitation in
applying those values directly.
In reality, the fuel or pyrolysis product is not consumed before ignition. However, due to the
mixing and burning controlling mechanism in FDS, the fuel was consumed or burnt before
the “ apparent ignition” . This can be verified from existing HRR curves before the “ apparent
ignition” (from Figure 7.1). Therefore the concentration and distribution of fuel were altered
by the combustion. The corresponding mixture fraction distribution was not the “ true”
distribution prior to the “ apparent ignition” . In fact, the obtained mixture fraction is the sum
of fuel generation rate (pyrolysis rate) and fuel consumption rate. Therefore applying the
mixture fraction as an ignition criterion, the effort from this consumption of fuel should be
investigated. One may obtain the mixture fraction without the “ apparent ignition” presence.
Alternatively, one may assume that the burning process prior to the apparent ignition did not
significantly alter the fuel distribution. Based on the relatively low HRR values before the
“ apparent ignition” (in Figure 7.1), as well as the fact that the ignition time (from the tests)
and the “ apparent ignition” time are very close above a radiative flux of 30 kW/m2 (shown
in Table 6.8), the latter assumption will apply with a certain degree of accuracy.
7.2 Comparison Between the Simulation and Experimental Data
To evaluate the applicability of these possible ignition criteria, comparison between the
simulation results and experimental data must be carried out. In this section the simulated
values are compared with experimental data from previous cone calorimeter tests as well as
the modelling and experimental data from other researchers.
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
195
7.2.1 Critical surface temperature
The simulated surface temperatures of the studied timbers at ignition range from around 326
to 343 oC for the various radiative fluxes. The test results for the mountain ash show a range
from 290 to 335 oC. The slightly higher prediction in simulations for the surface temperature
is not unusual (Zhou et al. 2002). Therefore the values of simulated critical surface
temperatures match with the cone calorimeter test data reasonably well.
When compared with the experimental results of others, the simulated values from current
study fall into an acceptable range. For example, Tzeng et al. (1990) reported a range from
580 to 660 K (307 to 387 oC) for maple, red oak, poplar and mahogany, under tested
radiative fluxes from 21 to 36 kW/m2. Yang et al. (2003) obtained surface temperatures for
ignition from 572 to 603 K (299 to 330 oC) for cherry wood with a thickness between 9 and
18 mm. In their tests, the radiation level ranges from 30 to 50 kW/m2. Under the same
radiative conditions, the surface temperatures for beech wood are from 563 to 593 K (290 to
320 oC).
7.2.2 Critical mass flux
The simulated critical mass fluxes for the mountain ash range from 1.0 to 1.6 g/m2⋅s under
radiation levels from 30 to 75 kW/m2. These values are lower than those measured in the
cone calorimeter tests. The release of moisture and other inert gases in the real fire scenarios
may contribute to the higher experimental values. As reviewed in Section 2.4.3, the
composition of the pyrolysis products of timber is very complex. There are still a number of
unknowns for these gaseous products. Therefore it is either impossible or unnecessary to
include all inert gases into current modelling.
When the simulated critical mass flux is compared with data from other research, it is still
within a range, 0.6 to 8 g/m2⋅s, which has been widely reported (see review in Section
2.4.3.3). Compared with the typical values for timber obtained by others, the current
simulated value is slightly lower. However the peak mass fluxes presented in Section 6.3.4
are similar to some values obtained by others, as shown below. For example, Yang et al.
(2003) reported 2 to 4 g/m2⋅s for the mass flux for cherry wood. A trend that shows a higher
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
196
critical mass flux under higher radiation levels (which was found in the current simulation)
was also reported by those researchers. Meanwhile the peak values that they provide are
around 6 to 7 g/m2⋅s under 30 to 50 kW/m2 radiation, which is slightly lower than the results
for the mountain ash but similar to that for pine wood in the current simulations, as shown
in Section 6.3.5. The reason for a lower critical mass flux range in the current simulation,
apart from the explanation in previous paragraph, may be the determination method adopted
here. Generally the simulated mass fluxes match other researchers’ experimental data
reasonably well.
7.2.3 Critical mixture fraction
Normally the concentration of pyrolysis products, especially the fuel concentration from the
pyrolysis, is very difficult to measure in the cone calorimeter test. There is very little data of
mixture fraction for timber materials in a cone calorimeter environment at this time.
Therefore only the simulation data that are available from others’ research are presented
here for some kind indirect comparison.
Tsai et al. (2001) performed simulations for the auto-ignition problem of PMMA in the cone
calorimeter. From their simulation results, the fuel concentration before ignition reached 0.3
to 0.5 g/g (mass fraction) at the centre point of the cone heater’ s bottom section. The
ignition criterion in this simulation was the critical mass flux. The kinetics used for the
pyrolysis calculation were obtained by a specified ignition temperature. It is noticed that
simulation results for the critical mixture fraction vary from different material properties as
well as the ignition criteria chosen. Therefore a direct comparison between values obtained
by Tsai et al. and that from the current study is not available.
Another example can be found in a numerical analysis carried out by Zhou et al. (2002),
again for PMMA. The physical problem was based on tests conducted on a FIST apparatus
and details have been presented in Chapter 2. In simulating ignition a spark plug was
positioned 0.5 to 1.5 mm above the PMMA plate. The mixture fraction between the solid
and the igniter ranged from 0.03 to 0.07 g/g (mass fraction) under 35 kW/m2 radiation and
1 m/s flow velocity. This range of values is slightly lower than current simulated data.
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
197
7.3 Special Consideration for the Critical Mixture Fraction
Acceptable agreement has been achieved between the simulated results and the
experimental data for the surface temperature and the mass flux. However no direct
comparison is available for the mixture fraction at this time. More theoretical analysis is
thus required for evaluating this possible ignition criterion.
7.3.1 Relationship between the critical mixture fraction and the LFL
It is found from Chapter 6 and the first section of current chapter that for mountain ash the
average mixture fraction values under various radiations at the ignition is about 0.020 and
0.027 g/g (at the spark plug and centre of surface). These average values will be checked for
possible relation to a LFL value which was based on the previously mentioned lean
flammability theory (Section 2.4.3.3). To perform this checking, the LFL value for timber
was calculated first. A comparison between the mixture fraction and the LFL of timber was
then carried out.
7.3.1.1 Calculation of LFL for timber
Although it is still difficult to obtain all details for the pyrolysis products of timber, there are
some simplified expressions suggested by other researchers (see Section 2.3.3). As provided
by Klose et al. (2000) from test results for three kinds of timbers, (Aspen, Beech, and
Larch), typically representative ratios for the major gases as mole fractions are: 42%, 25%,
8.3% and 24.7% for H2, CO, CH4, and CO2 respectively.
Applying the method shown in Section 2.4.3.3 (Equation (2.25)), the LFL of the pyrolyses
products from timber can be evaluated from the LFL of pyrolsate composites. It is also
known from the SFPE Handbook (Beyler and Hirschler 2002) that the LFLs for H2, CO, and
CH4 are 4.0%, 12.5%, and 5.0% in volume respectively. Therefore the lower flammable
limit at ambient temperature and pressure (room temperature and atmospheric press) for the
wood pyrolysis products, LFLpy, can be calculated as follows:
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
198
%06.7
5.1225
442
53.8
100 =
++=pyLFL
(7.2)
It is known that for a wide range of fuels containing carbon, hydrogen and oxygen, the
adiabatic flame temperature at the LFL is approximately 1600 K, except for hydrogen
(900 K) and carbon monoxide (1300 K) (Beyler 2002). Therefore the LFLs for H2, CO and
CH4 at an initial mixing temperature from pyrolysis (200 oC or 473 K for timber materials)
are calculated based on Equation (2.28) as follows:
For H2,
%8.2293900473900
4
293
473
)(,
)(,273473
2
2
=−−×=
−−
=KT
KTLFLLFL
HLFLf
HLFLfKK
(7.3)
For CO,
%3.1029313004731300
5.12
293
473
)(,
)(,273473
=−−×=
−−
=KT
KTLFLLFL
COLFLf
COLFLfKK
For CH4,
%3.429316004731600
%5
293
473
)(,
)(,273473
4
4
=−−×=
−−
=KT
KTLFLLFL
CHLFLf
CHLFLfKK
Applying Equation (2.25) again, the LFL for wood pyrolysis products at initial mixing
temperature of 200 oC is calculated as:
%2.5
3.1025
8.242
3.43.8
100 =
++=pyLFL
7.3.1.2 Converting the mixture fraction into volume fraction
Suppose a mixture fraction value at which ignition occurs is determined from the FDS
simulation. Since the “ mixing is combustion” mechanism is adopted in FDS, it represents a
similar situation as the piloted ignition in the cone calorimeter, i.e. ignition will occur when
a certain level of fuel concentration is reached at the location where the spark is present.
Therefore this mixture fraction can be related to a lean fuel concentration limit for ignition.
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
199
According to the definition of the mixture fraction, which was discussed in Chapter 6, a
volume fraction of the pyrolysis combustibles, VF, which is comparable to the LFL, can be
calculated by the mixture fraction (mass fraction as a ratio of gas phase fuel to total gas) via
the following equation.
)(1
)(
fractionmolar
MWZ
MWZ
MWZ
fractionvolumeVolumeVolume
VolumeVF
airFuel
Fuel
airFuel
FuelFuel
−+=
+=
(7.4)
where:
VFFuel: volume fraction of the gas phase fuel;
Z: mixture fraction, g/g;
MWFuel: molecular weight of the gas phase fuel, g/mol;
MWair: molecular weight of air, g/mol.
For the timber material, the molecular weight of its gas phase fuel, i.e. combustible gases in
the pyrolysis products, MWFuel, equals
MWFuel = 42%×2 (for H2) + 25%×28 (for CO) + 8.3%×16 (for CH4)
= 9.17 g/mol
Molecular weight of the air, MWair, equals
MWair = 21%×32 (for O2) + 79%×28 (for N2)
= 28.8 g/mol
Therefore the volume fraction for the gas phase fuel, VFFuel, is calculated from Equation
(7.4) as:
airFuelFuelFuel
FuelFuelFuel MWZMWZ
MWZVF
/)1(//
−+=
(7.5)
and then the volume fractions of fuel at the spark plug and the centre of solid surface can be
calculated from following:
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
200
%0.8
8.28973.0
17.9027.0
17.9027.0
%1.6
8.2898.0
17.902.0
17.902.0
,
,
=+
=
=+
=
surfaceFuel
plugFuel
VF
VF
It can be seen that the volume fractions of pyrolysis products, which were calculated from
the mixture fraction values, are slightly above the lower flammable limit (5.2% volume) for
the timber pyrolysis products. This is an indication that the critical mixture fraction is
related to the lower flammable limit to some degree. The value of the critical mixture
fraction for the timber can be calculated from Equation (7.4) as 0.017 g/g, which is just
below the simulated minimum critical mixture fraction value shown in Table 7.1.
Calculation results for the LFL and VF may vary when adopted pyrolysis gas composition
changed. Reported gas composition varies depending on species of timber, decomposition
environment, and testing/measuring equipments, etc. (Chembukulam et al. 1981; Lucon
1997; Klose et al. 2000). However the relationship between the LFL and VF is maintained.
For example, by using a set of ratios provided by Yao et al (Yao et al. 2006), as
CH4:H2:CO:CO2 = 24.0:25.8:38.3:11.9 (v/v %), following result can be obtained. At 200 oC,
LFLpy is 5.0%, and VFFuel,surf calculated from the mixture fraction of 0.027 g/g is 5.1%,
which is still slightly above the LFL.
7.3.2 Discussion for the suggested criterion
Among the three criteria for ignition of solid materials, critical surface temperature may be
normally considered the most useful since it can be conveniently related to fire spread
(Atreya 1983). Some models use this criterion reasonably well, but only in certain situations.
Critical surface temperature is an empirical criterion and there is plenty of experimental data
available for modelling and validating. However the surface temperature at ignition will be
affected by several factors, such as external radiation and oxidizer flow conditions, which
are still not fully understood. Meanwhile, as indicated in the literature review, there is great
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
201
variation for the reported value of this parameter. This makes it difficult to adopt test data
and widely apply it, especially in complex gas flow conditions.
Critical mass flux, as a primary index for pyrolysis progress, is another important criterion
for ignition. It is a useful one, specially under some given flow conditions, such as certain
quasi-steady periods, and has been suggested by a number of researchers (Bamford et al.
1946; Drysdale and Thomson 1989). The limitation of this parameter as a criterion is that
the critical mass flux varies under various radiation and air flow conditions. An example of
varied critical mass flux under varying radiative fluxes has been given in Table 6.7 which
has been also reported by other researchers (Zhou et al. 2002; Yang et al. 2003).
The mixture fraction is the third potential candidate as the criterion of wood ignition in a
cone calorimeter environment. As discussed early in this section, relationship between this
criterion and the lower flammable limit has been found. This enables the fourth element in
the combustion tetrahedron (discussed in Section 2.4.3) to be met. While a spark exists in a
piloted ignition and fresh air was supplied into the space above the specimen in a cone
calorimeter environment, satisfying of the fourth element at the spark plug will lead to
ignition occurring immediately. Considering other two parameters, critical surface
temperature and mass flux, they just satisfy one leg of the tetrahedron separately and then
do not necessarily result in ignition. From previous simulation results, it is observed that the
mixture fraction criterion works quite well under either various radiation levels (shown in
Table 7.1) or various gas velocities (shown in Table 6.7). Therefore it may be expected that
this mixture fraction will be useful for prediction of ignition under complicated conditions.
There are still much to be done before the mixture fraction becomes a robust and solid
criterion. Most will come from obtaining appropriate pyrolysis information, especially the
kinetics, and validating it under a wide range of environment conditions. On the other hand,
the computed LFL or mixture fraction should be ideally validated through measurement of
fuel concentration at the location of the igniter (spark plug) and at the time of ignition in a
cone calorimeter environment. While it is difficult to obtain such measurement with the
available technology and resources, further improvement for the numerical scheme, which
was used alternatively to obtain the fuel concentration at the igniter, should be carried out.
The most critical one may be improvement of the ignition controlling mechanism in the
FDS model. The “ mixed is burnt” mechanism may affect the accuracy of computation of the
Chapter 7 Criteria for Ignition and Discussion ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
202
mixture fraction. Meanwhile this mechanism makes FDS suitable to model the piloted
ignition where the gas phase time delay can be ignored, according to the basic analysis for
ignition controlling mechanism performed in Section 2.4. When the real mixture fraction
(generated from pyrolysis of solid) can be obtained accurately, the ideal methodology to
predict ignition phenomena will rely on a set of criteria, including the critical mixture
fraction, local temperature and oxygen concentration. While the latter two conditions are
relatively easy to be verified, this approach of ignition prediction can be applied onto wider
and complex fire scenarios, such as non-piloted ignition in the cone calorimeter and flame
spread over non-adjacent objects.
In summary, critical mixture fraction is a potentially useful ignition criterion. Compared
with the other two commonly adopted criteria, critical surface temperature and critical mass
flux, it has several significant advantages. First, it is able to present the gas phase fuel
distribution and predict ignition based on the LFL theory in a piloted cone calorimeter
condition. Secondly, it is relatively stable in various radiation and gas flow conditions.
However, to apply this criterion relies on accurate and applicable kinetics for the studied
material. This remains a major task not just for FDS but also for all pyrolysis models with
pyrolysis flux calculation function since the available kinetic information of various
materials is still very limited. Suggested approach from current research shows a possibility
for all the pyrolysis models to produce a more accurate prediction of ignition and wider
application of modelling in complicated fuel configurations and environments.
203
®¯°±²³´!µ®¯°±²³´!µ®¯°±²³´!µ®¯°±²³´!µ ®H¶·D®j¸h¹»º¼L¶M·Gº°·½®H¶·D®j¸h¹»º¼L¶M·Gº°·½®H¶·D®j¸h¹»º¼L¶M·Gº°·½®H¶·D®j¸h¹»º¼L¶M·Gº°·½´³®¶M¾I¾D³&·½¿°'²¼L¶M·GºDÀk¶M´ÁÀG¹´²¯³´´³®¶M¾I¾D³&·½¿°'²¼L¶M·GºDÀk¶M´ÁÀG¹´²¯³´´³®¶M¾I¾D³&·½¿°'²¼L¶M·GºDÀk¶M´ÁÀG¹´²¯³´´³®¶M¾I¾D³&·½¿°'²¼L¶M·GºDÀk¶M´ÁÀG¹´²¯³´´³º³°Q´®j¯´³º³°Q´®j¯´³º³°Q´®j¯´³º³°Q´®j¯
8.1 Conclusion
Quantitative analysis of the pyrolysis and ignition processes of the studied furnishing
materials has been achieved to varying degrees. By improving the testing and computational
scheme for the pyrolysis kinetics, the obtained “ effective” kinetics have been applied into a
comprehensive CFD model, FDS, and the pyrolysis process of the studied timber in a cone
calorimeter has been simulated with reasonable accuracy. Through the modelling, a gas
phase fuel concentration parameter, the mixture fraction, has been monitored in a cone
calorimeter environment. Validating the mixture fraction at ignition against with lower
flammable limit (LFL), it is found that the critical mixture fraction is a potential and robust
ignition criterion. It has the advantage that it can be applied to complicated environments.
Different decomposition behaviours of the studied materials under various heating
conditions have been observed. The testing method and computational scheme of TGA for
acquiring the kinetics were then improved to reflect the different decomposition behaviours.
By applying the high heating rate combined with the linear shifting relationship of the
reaction rate, the obtained “ effective” kinetics are capable to describe the decomposition
under environmental conditions similar to real fires. These observations and improvements
are also useful and applicable for modelling decomposition of some other complex solid
materials by an overall scheme. However how to model the sub-reactions from multiple
composites in various heating conditions remains as a challenge.
The pyrolysis and ignition processes for the materials have been experimentally studied in a
cone calorimeter. The major measured parameters matched the experimental data of others.
Some interesting phenomena related to charring were observed. A small peak value was
observed in the mass loss rate curves before ignition for charring materials under lower
radiation levels. This seldom reported phenomenon matches Atreya’ s theoretical analysis
that the thicker char layer, if formed before ignition, will slow down the pyrolysis rate. A
Chapter 8 Conclusions and Recommendations for Further Research ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
204
data processing technique for the gas yield measurements has been improved. The errors in
measurement have been reduced by an integral method.
A CFD model, FDS, which is based on Atreya’ s one-step overall pyrolysis model, was used
for performing numerical simulations. By applying the kinetics from the scheme, which this
researcher modified, the simulated major ignition parameters matched the experimental data
reasonably well. A gas phase parameter, the mixture fraction, was monitored at desired
locations. This successfully extended the application areas of FDS into detailed pyrolysis
and ignition processes with engineering accuracy.
The critical mixture fractions for mountain ash under various conditions were found
approximately 0.020 g/g at the plug or 0.027 g/g at the center of solid surface. The
calculated values are just above the lower flammable limit (LFL) computed for the timber
(0.017 g/g). This establishes a relationship between the critical mixture fraction and lean
fuel concentration limit for ignition. The mixture fraction is thus suggested by the author to
be a suitable gas phase ignition criterion. Compared with other commonly accepted criteria
this parameter is able to predict ignition in the simulated environment more accurately and
reliably. The existed theoretical basis makes the mixture fraction as a robust criterion and
indicates a wider application.
8.2 Further Research Recommendations
There are many aspects of the modelling of pyrolysis and ignition of solid fuels that are
required for further research. Generally, two types of tasks still remain: (1) to obtain more
basic details about the pyrolysis and ignition processes and to model them effectively, (2) to
validate the ignition criteria to apply fire models onto wider and more complex geometries.
For example, complex pyrolysis products for commonly used materials are still not fully
understood; some phenomena, such as crack and shrink for timber and melt for PU foam,
must be taken into consideration in modelling, etc. This research work has shown an
engineering approach to combining more phenomena into modelling and to generate more
accurate prediction based on a solid theory of ignition controlling mechanism. To perfect
this approach, it is recommended that the further development of pyrolysis and ignition
should involve the following aspects:
Chapter 8 Conclusions and Recommendations for Further Research ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
205
1. The carrying out of TG type tests under even higher heating rate conditions. Since it
has been found that there are different thermal decomposition behaviours and
kinetics in low and high heating rate ranges, extending the heating rate to those
closer to real fire conditions is naturally the next step. Meanwhile such test results
can be used to validate the shifting pattern in that higher heating rate range.
2. More details of wood pyrolysis should be included in the current pyrolysis model.
These details may include phenomena such as shrinkage and cracking. While the
current simulation has not included a relationship between pyrolysis and surface
oxidant concentration that theoretically exists, further research may be necessary for
extending the modelling ability of FDS on such phenomena.
3. Measurement of fuel concentration at spark plug location in a cone calorimeter
environment is desired. Such experimental data will approve the relationship
between the critical mixture fraction and the LFL directly.
4. In FDS, some type of modelling should be performed for investigating individual
pyrolysis compounds. More and more kinetic parameters from multi-reaction or sub-
reaction of complex decompositions are being studied by many researchers to obtain
detailed understanding of pyrolysis and combustion, especially for toxic gaseous
products. However, such complicated reactions are still unable to be modelled in
most of current fire models. Such improved simulations will be helpful to evaluate
the mixture fraction criterion proposed by this author.
5. Another major improvement for the FDS model may be the modification of the
current ignition controlling mechanism. As discussed in Chapter 6 and 7, functions
such as to “ turn off” the combustion and to control reaction rate by local oxygen
concentration may be helpful to obtain a more accurate and meaningful mixture
fraction generated from pyrolysis.
6. Apply the suggested mixture fraction criterion to other environments. Combined
with other criteria, like local gas temperature, one may predict ignition in more
complicated scenarios, such as non-piloted ignition and non-adjacent flame spread.
206
ÂQûÃ\Ä&ÅÆ=ÇÈÉÂÊ=ËRÌÍ ËÄÏÎ=ËÂQûÃ\Ä&ÅÆ=ÇÈÉÂÊ=ËRÌÍ ËÄÏÎ=ËÂQûÃ\Ä&ÅÆ=ÇÈÉÂÊ=ËRÌÍ ËÄÏÎ=ËÂQûÃ\Ä&ÅÆ=ÇÈÉÂÊ=ËRÌÍ ËÄÏÎ=ËÐÎÄ|Ë)ËÇBÅDÌ¿ÎÎÄ|Ë)ËÇBÅDÌ¿ÎÎÄ|Ë)ËÇBÅDÌ¿ÎÎÄ|Ë)ËÇBÅDÌ¿ÎÂÅÆEÑ&ÄÎ)ÒÓË?ÎÂÅÆEÑ&ÄÎ)ÒÓË?ÎÂÅÆEÑ&ÄÎ)ÒÓË?ÎÂÅÆEÑ&ÄÎ)ÒÓË?Î
Listed in this chapter are whole set of major experimental settings and testing results of
TGA tests for the studied materials.
Table A1 to A6 show the settings and raw results for each testing material. For the dynamic
(DYN) tests, the heating rate was quoted and a constant stable environmental temperature
was listed for the isothermal (ISO) tests.
The test codes were given as following rules:
TG XX mm
where:
TG: the TGA test;
XX: type of material, CT: cotton;
CP: cotton and polyester;
PI: pine;
MA: mountain ash;
SA: Stamina PU foam;
SD: Standard PU foam;
mm: test order number.
The “ -s” in the mass column for the timbers represents saw dust, at a size of around 100
meshes. The parameter WT refers to sample weight for tested materials.
Table A1 TGA tests for the cotton and polyester
TGCP01 5.27 air DYN 10 355 0.42 69.2 431 0.42 33.0TGCP02 5.35 air DYN 20 365 0.94 68.6 449 0.73 34.0TGCP03 4.72 air DYN 30 394 0.74 77.1 489 1.00 38.8TGCP04 3.33 air DYN 30 376 0.73 71.4 451 0.77 37.8TGCP05 2.58 air DYN 50 412 0.73 75.5 485 1.08 40.6TGCP06 3.11 air DYN 100 454 1.85 76.7 559 1.76 36.2TGCP07 4.50 air DYN 200 476 7.60 66.7 560 3.30 33.6TGCP08 2.85 N2 DYN 20 394 0.43 66.4 461 0.40 36.0TGCP09 3.53 N2 DYN 30 372 0.74 66.8 462 0.67 30.7
Heating rate
(K/min)
Tmax,1
(oC)MLRmax,1
(mg/min)Test code
Working gas
Test type
Mass (mg)
WTmax,1
(%)Tmax,2
(oC)MLRmax,2
(mg/min)WTmax,2
(%)
Appendix A TGA Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
207
Table A2 TGA tests for the cotton
Table A3 TGA tests for the mountain ash
TGMA01 4.81 Air DYN 5 320 1.60 59.5TGMA02 3.36 Air DYN 10 339 0.48 54.0TGMA03 2.16 Air DYN 20 343 10.50 56.1TGMA04 3.14 Air DYN 30 353 52.50 56.4TGMA05 3.59 Air DYN 50 470 76.90 50.0TGMA06 5.60 Air DYN 100 505 433.70 47.2TGMA07 4.32 Air DYN 200 516 1,272.90 45.6TGMA08 2.65 N2 DYN 10 367 0.33 41.0TGMA09 3.00 N2 DYN 30 369 1.21 45.0TGMA10 2.70-s Air DYN 10 380 3.29 50.1TGMA11 4.37-s Air DYN 30 411 40.90 49.6TGMA12 3.40-s Air DYN 75 437 188.00 45.7TGMA13 4.81-s N2 DYN 10 409 4.10 48.3TGMA14 4.13-s N2 DYN 30 438 30.50 45.2TGMA15 2.65 Air ISO 320 0.82 90.1TGMA16 2.48 Air ISO 340 1.39 85.0TGMA17 2.57 Air ISO 350 1.61 85.1TGMA18 2.44 Air ISO 375 1.96 80.5
Test code
Working gas
Test type
Mass (mg)
Heating rate
(K/min)
Constant Temp. (oC)
Tmax
(oC)MLRmax
(mg/min)WTmax
(%)
TGCT01 3.49 air DYN 10 367 1.10 43.5TGCT02 6.99 air DYN 10 368 1.30 50.2TGCT03 5.25 air DYN 20 377 7.20 35.8TGCT04 6.94 air DYN 20 370 3.67 42.8TGCT05 6.10 air DYN 30 385 10.80 36.7TGCT06 4.55 air DYN 50 447 329.00 36.6TGCT07 4.36 air DYN 75 476 372.80 42.8TGCT08 3.71 air DYN 100 500 524.00 44.8TGCT09 3.95 air DYN 150 520 1085.20 43.0TGCT10 3.88 air DYN 200 540 1731.00 46.0TGCT11 3.95 N2 DYN 10 382 3.60 46.7TGCT12 5.18 N2 DYN 20 400 1.60 52.9TGCT13 7.40 N2 DYN 30 383 3.61 44.6TGCT14 4.07 N2 DYN 75 503 245.40 47.8TGCT15 3.73 N2 DYN 150 535 799.50 49.1TGCT16 6.09 air ISO 335 0.38 95.2TGCT17 5.73 air ISO 340 0.43 95.1TGCT18 6.36 air ISO 350 0.69 95.3TGCT19 6.09 air ISO 360 1.20 90.0TGCT20 6.00 air ISO 380 2.96 83.7
Working gas
Test code
Test type
Mass (mg)
Heating rate
(K/min)
Constant Temp. (oC)
Tmax
(oC)MLRmax
(mg/min)WTmax
(%)
Appendix A TGA Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
208
Table A4 TGA tests for the pine
TGPI01 4.70 Air DYN 10 359 0.77 52.7TGPI02 7.50 Air DYN 20 381 1.70 41.1TGPI03 2.20 Air DYN 30 390 0.72 37.7TGPI04 3.48-s Air DYN 30 425 1.08 48.0TGPI05 3.83-s Air DYN 75 449 2.79 46.4TGPI06 1.40 N2 DYN 10 364 2.30 44.9TGPI07 3.20 N2 DYN 20 410 0.72 40.8TGPI08 3.60 N2 DYN 30 391 1.20 39.1TGPI09 3.76-s N2 DYN 30 452 0.95 50.8
Heating rate
(K/min)
Tmax
(oC)MLRmax
(mg/min)Test code
Working gas
Test type
Mass (mg)
WTmax
(%)
Table A5 TGA tests for the stamina PU foam
TGSA01 2.37 air DYN 10 284 0.26 76.4 352 0.22 23.9TGSA02 3.10 air DYN 20 283 0.72 78.2 348 0.50 29.4TGSA03 2.84 air DYN 30 287 0.76 80.8 370 0.71 20.8TGSA04 1.50 air DYN 50 379 0.50 76.7 457 0.57 26.3TGSA05 1.48 air DYN 100 397 0.89 74.8 484 0.96 28.2TGSA06 0.80 N2 DYN 20 341 0.19 41.8TGSA07 1.76 N2 DYN 30 294 0.57 77.1 354 0.46 28.8TGSA08 1.25 N2 DYN 50 398 0.24 76.4 498 1.10 25.1TGSA09 1.97 N2 DYN 100 391 0.67 82.7 517 3.28 29.4
Heating rate
(K/min)
Tmax,1
(oC)MLRmax,1
(mg/min)Test code
Working gas
Test type
Mass (mg)
WTmax,1
(%)Tmax,2
(oC)MLRmax,2
(mg/min)WTmax,2
(%)
Table A6 TGA tests for the standard PU foam
TGSD01 1.19 air DYN 5 277 0.12 59.2TGSD02 1.92 air DYN 10 286 0.33 62.6 318 0.21 27.3TGSD03 2.44 air DYN 20 293 0.72 69.3 344 0.51 19.9TGSD04 2.44 air DYN 30 297 0.90 73.3 361 0.75 20.9TGSD05 1.50 air DYN 50 381 0.63 71.3 453 0.54 21.2TGSD06 0.77 air DYN 100 416 0.74 64.6 456 0.83 27.1TGSD07 1.04 air DYN 150 423 1.43 64.4 461 1.61 31.1TGSD08 0.91 air DYN 200 474 1.78 26.7TGSD09 1.39 N2 DYN 20 311 0.22 78.4 389 0.47 28.4TGSD10 1.24 N2 DYN 30 313 0.31 76.0 359 0.42 38.7TGSD11 1.78 N2 ISO 250 0.10 98.5TGSD12 2.14 N2 ISO 300 0.56 96.7TGSD13 1.49 N2 ISO 300 0.41 95.5
Test code
Working gas
Test type
Mass (mg)
Heating rate
(K/min)
Constant Temp. (oC)
Tmax,1
(oC)MLRmax,1
(mg/min)WTmax,1
(%)Tmax,2
(oC)MLRmax,2
(mg/min)WTmax,2
(%)
209
ÔQÕ»Õ\Ö&×Ø=ÙÚ;ÛkÜOÝÞM×JÖÝwÔßkÞàÙBáIÖâÖ&à âÖÏã=âÔQÕ»Õ\Ö&×Ø=ÙÚ;ÛkÜOÝÞM×JÖÝwÔßkÞàÙBáIÖâÖ&à âÖÏã=âÔQÕ»Õ\Ö&×Ø=ÙÚ;ÛkÜOÝÞM×JÖÝwÔßkÞàÙBáIÖâÖ&à âÖÏã=âÔQÕ»Õ\Ö&×Ø=ÙÚ;ÛkÜOÝÞM×JÖÝwÔßkÞàÙBáIÖâÖ&à âÖÏã=âÐãÖ|â)âÙB×Dä¿ããÖ|â)âÙB×Dä¿ããÖ|â)âÙB×Dä¿ããÖ|â)âÙB×Dä¿ãÔQ×ØEàÖÏã)åßâ?ãÔQ×ØEàÖÏã)åßâ?ãÔQ×ØEàÖÏã)åßâ?ãÔQ×ØEàÖÏã)åßâ?ã
Listed in this appendix are whole set of major experimental settings and testing results for
the tested materials in the cone calorimeter.
Table B1 to B6 show the settings and results from the tests for each tested material,
including those tests for the surface temperature measuring, individually.
The test codes were given as following:
For the woods and polyurethane foams,
X-mm-nn-Y-Z
X is the type of material, where:
MA: mountain ash;
PI: Pine;
SD: Standard PU foam;
SA: Stamina PU foam
mm is the sample thickness, mm
nn is external radiation, kW/m2
Y is the pilot method, where:
P: piloted; and
N: non-piloted.
Z is the order number.
For the fabrics,
XX-0-nn-Y-Z
where XX is material,
CP: cotton and polyester; and
CT: cotton
0 means single layer of the fabrics tested.
All lefts have same meanings as above.
In the result columns, following parameters are used:
tig: ignition time, s
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
210
HRRmax: maximum heat release rate, kW/m2
THR: total heat released, MJ/m2
EHCave: average effective heat of combustion, MJ/kg
Generally, “ Smouldering” represents a status of combustion, smouldering. “ NC” means no
combustion, includes the smouldering, observed.
Table B1 Cone calorimeter tests for the mountain ash
Thickness Radiation tig HRRmax THR EHCave
(mm) (kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)
MA3230P1 32 30 Y 133 155.9 220.2 13.0MA3230P2 32 30 Y 154 159.6 212.5 5.2MA3230P3 32 30 Y 248 150.2 239.0 10.8MA3230P4 32 30 Y 77 152.4 208.6 11.2MA3230N1 32 30 N 85.3 129.8 10.6 SmouldingMA3230N2 32 30 N 32.2 103.8 11.4 SmouldingMA3240P1 32 40 Y 52 178.1 219.5 11.2MA3240P2 32 40 Y 46 176.7 210.9 11.6MA3240P3 32 40 Y 61 171.9 230.0 11.5MA3240N1 32 40 N 68 169.1 212.3 11.4MA3240N2 32 40 N 60 180.1 206.2 11.7MA3240N3 32 40 N 56 183.2 210.8 11.6MA3250P1 32 50 Y 23 234.3 204.1 11.0MA3250P2 32 50 Y 26 226.4 205.9 11.5MA3250P3 32 50 Y 29 223.4 209.6 11.9MA3250P4 32 50 Y 30 261.3 210.0 11.7MA3250N1 32 50 N 29 224.1 202.9 11.0 MA3250N2 32 50 N 37 201.3 204.1 11.2MA3250N3 32 50 N 46 194.7 253.2 12.1MA3250N4 32 50 N 45 203.5 203.7 11.5MA3250N5 32 50 N 24 243.6 213.8 11.7MA3275N1 32 75 N 13 353.5 259.9 13.2MA3275N2 32 75 N 10 345.8 246.4 12.6MA3275N3 32 75 N 9 326.0 250.0 12.9MA1430P1 14 30 Y 89 186.4 94.2 11.6MA1430P2 14 30 Y 111 176.3 98.1 11.3MA1430P3 14 30 Y 95 192.6 102.5 11.7MA1450P1 14 50 Y 30 191.9 100.8 12.1MA1450P2 14 50 Y 27 198.3 110.5 12.3MA1450P3 14 50 Y 31 204.9 98.7 12.5MA1450N1 14 50 N 45 202.3 106.1 12.2MA1450N2 14 50 N 51 197.7 112.0 12.0MA1450N3 14 50 N 69 212.8 103.8 12.8
NoteTest codePiloted
Settings Results
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
211
Table B2 Cone calorimeter tests for the pine wood
Thickness Radiation tig HRRmax THR EHCave
(mm) (kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)PI3230P1 32 30 Y 129 160.4 191.4 12.4PI3230P2 32 30 Y 92 155.2 196.7 12.5PI3230P3 32 30 Y 101 151.0 182.2 12.4PI3230P4 32 30 Y 63 167.3 186.5 12.3PI3230N1 32 30 N 150.0 120.8 7.9 SmouldingPI3230N2 32 30 N 39.1 86.2 5.8 SmouldingPI3230N3 32 30 N 30.4 64.6 4.4 SmouldingPI3240P1 32 40 Y 34 158.8 161.2 12.0PI3240P2 32 40 Y 37 163.5 160.2 11.5PI3240P3 32 40 Y 34 162.7 168.5 11.8PI3240N1 32 40 N 92 158.8 160.8 12.5PI3240N2 32 40 N 40 163.1 179.3 12.3PI3240N3 32 40 N 65 162.1 165.1 11.8PI3250P1 32 50 Y 21 180.3 175.5 11.7PI3250P2 32 50 Y 21 152.7 191.3 12.4PI3250P3 32 50 Y 18 171.8 176.1 11.6PI3250P4 32 50 Y 17 191.2 182.9 11.9PI3250N1 32 50 N 37 180.1 158.3 12.0 PI3250N2 32 50 N 29 179.3 164.6 12.0PI3250N3 32 50 N 31 184.0 209.7 12.1PI3250N4 32 50 N 25 199.5 176.8 12.3PI3275N1 32 75 N 7 233.1 162.2 12.0PI3275N2 32 75 N 8 245.0 173.8 12.3PI3275N3 32 75 N 9 228.7 169.5 12.6
Settings Results
NoteTest codePiloted
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
212
Table B3 Cone calorimeter tests for the cotton and polyester
Radiation tig HRRmax THR EHCave
(kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)CP020P1 1 20 Y 34 199.6 7.1 16.6CP020P2 1 20 Y 34 210.3 7.0 17.1CP020P3 1 20 Y 36 210.3 7.0 17.4CP030P1 1 30 Y 17 249.6 7.4 16.7CP030P2 1 30 Y 16 275.2 7.6 16.3CP030P3 1 30 Y 19 257.4 7.6 17.2CP030N1 1 30 N 18 246.6 7.2 16.3CP030N2 1 30 N 17 249.5 7.0 17.3CP030N3 1 30 N 20 254.0 7.3 16.9CP040P1 1 40 Y 10 296.6 7.6 17.2CP040P2 1 40 Y 9 310.5 7.7 17.9CP040P3 1 40 Y 7 320.8 7.2 17.3CP040N1 1 40 N 12 271.4 7.8 16.8CP040N2 1 40 N 13 270.8 7.1 17.4CP040N3 1 40 N 13 296.6 7.3 17.9CP050P1 1 50 Y 6 361.9 7.7 16.3CP050P2 1 50 Y 6 344.0 7.8 17.4CP050P3 1 50 Y 7 360.1 7.5 16.6CP050N1 1 50 N 5 326.8 7.6 16.5CP050N2 1 50 N 9 305.5 7.5 17.8CP050N3 1 50 N 9 322.6 7.6 17.7CP075N1 1 75 N 5 486.2 8.6 17.2CP075N2 1 75 N 5 413.9 8.8 16.2CP075N3 1 75 N 6 433.5 8.5 17.1
ResultsTest code
PilotedLayer
Settings
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
213
Table B4 Cone calorimeter tests for the cotton fabric
Radiation tig HRRmax THR EHCave
(kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)CT020P1 1 20 Y 54 137.8 5.9 14.5CT020P2 1 20 Y 67 143.9 6.3 15.8CT020P3 1 20 Y 56 148.7 6.1 14.8CT020N1 1 20 N 36.3 3.2 5.0 SmouldingCT020N2 1 20 N 3.3 3.1 0.4 SmouldingCT020N3 1 20 N 29.4 3.4 4.8 SmouldingCT030P1 1 30 Y 24 214.8 7.0 14.2 CT030P2 1 30 Y 21 233.5 6.8 14.6 CT030P3 1 30 Y 22 227.6 7.4 15.5 CT030N1 1 30 N 26 193.3 6.5 14.5CT030N2 1 30 N 24 203.1 7.0 14.8CT030N3 1 30 N 35 134.6 5.2 17.8CT040P1 1 40 Y 16 257.6 7.6 15.2CT040P2 1 40 Y 14 290.7 7.4 15.6CT040P3 1 40 Y 16 231.1 7.5 14.6CT040N1 1 40 N 21 189.5 5.5 15.2CT040N2 1 40 N 16 251.4 7.4 14.9CT040N3 1 40 N 18 232.5 6.1 15.1CT050P1 1 50 Y 13 272.0 7.5 15.9CT050P2 1 50 Y 11 336.2 7.0 16.0 CT050P3 1 50 Y 10 305.6 7.8 15.6CT050N1 1 50 N 13 277.1 7.6 15.9CT050N2 1 50 N 12 293.2 7.3 15.4CT050N3 1 50 N 11 275.8 7.5 15.5CT075N1 1 75 N 7 402.0 8.4 13.1CT075N2 1 75 N 8 387.9 8.6 15.0CT075N3 1 75 N 7 412.5 7.9 14.6
Results
NoteTest codePilotedLayer
Settings
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
214
Table B5 Cone calorimeter tests for the standard PU foam
Thickness Radiation tig HRRmax THR EHCave
(mm) (kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)SD5020P1 50 20 Y 12 398.7 33.1 26.7SD5020P2 50 20 Y 5 366.5 33.3 25.8SD5020P3 50 20 Y 6 386.7 34.5 27.3SD5020P4 50 20 Y 9 395.9 33.7 26.1SD5020N1 50 20 N 4.3 0.2 1.0 SmouldingSD5020N2 50 20 N 6.4 2.0 1.6 SmouldingSD5020N3 50 20 N 8.7 3.1 2.4 SmouldingSD5030P1 50 30 Y 3 426.2 32.0 25.7 SD5030P2 50 30 Y 4 499.1 33.1 26.4 SD5030P3 50 30 Y 3 419.7 32.9 27.1 SD5030N1 50 30 N 4.0 0.3 0.3 SmouldingSD5030N2 50 30 N 27 485.5 30.0 27.5SD5030N3 50 30 N 18 457.1 31.1 26.5SD5030N4 50 30 N 18 454.2 30.1 26.7SD5040P1 50 40 Y 3 459.1 32.2 25.7SD5040P2 50 40 Y 2 471.6 32.5 26.5SD5040P3 50 40 Y 1 448.2 33.3 26.6SD5040N1 50 40 N 14 477.7 29.7 25.7SD5040N2 50 40 N 6 425.3 32.5 26.9SD5040N3 50 40 N 4 436.9 32.2 26.0SD5050P1 50 50 Y 2 515.5 35.3 26.4SD5050P2 50 50 Y 1 481.6 34.6 27.2 SD5050P3 50 50 Y 2 486.1 32.9 26.8SD5050N1 50 50 N 8 497.8 33.6 27.1SD5050N2 50 50 N 4 546.1 39.1 26.3SD5050N3 50 50 N 3 503.8 33.6 26.2SD5050N4 50 50 N 3 507.2 33.5 26.5SD5075N1 50 75 N 1 656.7 37.5 28.4SD5075N2 50 75 N 1 627.7 35.0 27.3SD5075N3 50 75 N 2 633.0 36.2 28.1
Results
NoteTest codePiloted
Settings
Appendix B Cone Calorimeter Test Settings and Results ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
215
Table B6 Cone calorimeter tests for the stamina PU foam
Thickness Radiation tig HRRmax THR EHCave
(mm) (kW/m2) (s) (kW/m2) (MJ/m2) (MJ/kg)
SA5010P1 50 10 Y 38 126.4 6.2 22.2SA5010P2 50 10 Y 45 168.1 6.9 22.2SA5010P3 50 10 Y 41 145.2 6.7 22.6SA5020P1 50 20 Y 7 389.5 43.1 25.2SA5020P2 50 20 Y 6 329.8 43.7 25.9SA5020P3 50 20 Y 11 367.1 43.0 25.6SA5020N1 50 20 N 1.4 0.0 0.1 NCSA5020N2 50 20 N 6.1 1.4 1.3 SmouldingSA5020N3 50 20 N 6.0 0.3 1.3 SmouldingSA5030P1 50 30 Y 8 433.2 44.4 25.9 SA5030P2 50 30 Y 6 412.7 42.9 25.6 SA5030P3 50 30 Y 4 396.8 44.8 25.7 SA5030N1 50 30 N 32 440.7 41.1 26.1SA5030N2 50 30 N 31 386.4 41.1 25.9SA5030N3 50 30 N 17 404.7 43.3 24.9SA5040P1 50 40 Y 5 391.5 46.3 25.9SA5040P2 50 40 Y 4 438.7 43.7 25.4SA5040P3 50 40 Y 3 410.8 44.3 25.5SA5040N1 50 40 N 21 501.3 45.0 27.3SA5040N2 50 40 N 19 491.1 42.1 25.2SA5040N3 50 40 N 11 391.4 43.7 25.6SA5050P1 50 50 Y 1 386.2 43.9 25.1SA5050P2 50 50 Y 1 393.8 43.6 25.2 SA5050P3 50 50 Y 1 442.7 44.1 25.0SA5050N1 50 50 N 12 497.3 46.4 25.8SA5050N2 50 50 N 15 398.3 42.6 25.1SA5050N3 50 50 N 4 493.1 45.1 25.7SA5075N1 50 75 N 3 579.7 44.8 25.2SA5075N2 50 75 N 3 573.1 43.7 25.5SA5075N3 50 75 N 2 584.3 44.5 25.9SA2530P1 25 30 Y 5 495.4 25.4 27.0SA2530P2 25 30 Y 6 349.9 18.6 26.9 SA2530P3 25 30 Y 12 398.4 23.5 25.3SA2530N1 25 30 N 63 411.1 19.6 28.1SA2530N2 25 30 N 45 423.4 20.4 26.3SA2530N3 25 30 N 31 463.0 24.2 25.7SA2550P1 25 50 Y 4 617.1 24.5 26.8SA2550P2 25 50 Y 2 576.6 20.9 21.6SA2550P3 25 50 Y 5 539.4 21.4 22.6SA2550N1 25 50 N 4 578.1 26.8 27.3SA2550N2 25 50 N 4 621.8 25.3 27.0SA2550N3 25 50 N 6 609.4 25.9 27.3
Test code
Results
NotePiloted
Settings
216
æQç»ç\è&éê=ëìæQç»ç\è&éê=ëìæQç»ç\è&éê=ëìæQç»ç\è&éê=ëì íí íí\î.æéïæðhñóòëiòæQéê íHôõDçGö÷øæ'÷ëôé ôùî.æéïæðhñóòëiòæQéê íHôõDçGö÷øæ'÷ëôé ôùî.æéïæðhñóòëiòæQéê íHôõDçGö÷øæ'÷ëôé ôùî.æéïæðhñóòëiòæQéê íHôõDçGö÷øæ'÷ëôé ôùæ'÷úè|ñæüû òæ'÷úè|ñæüû òæ'÷úè|ñæüû òæ'÷úè|ñæüû ò-çñ%úôMðñòóëýòDõ+ôê=è&ðçñ%úôMðñòóëýòDõ+ôê=è&ðçñ%úôMðñòóëýòDõ+ôê=è&ðçñ%úôMðñòóëýòDõ+ôê=è&ð
All the non-dimensional parameters marked with [*] in this appendix have the same
meaning with the corresponding dimensional parameters shown in the Notation.
For a physical heating model described in Section 3.2, following assumption is applied to
further simplify the boundary conditions. The thermal conductivity is proportional to the
density, i.e., *λ = *ρ k, where k = constant = λ∞/ρ∞. Therefore, the non-dimensional basic
equations for the pyrolysis model of a wood material can be written as follows:
Energy:
∂∂
∂∂=
∂∂
∗
∗
∗∗
∗
yT
ytT ** ρρ (C. 1)
Decomposition: ( ) ∗∗−∗∗∗∗
∗
−−=∂∂ TEeA
t/δρρ
(C. 2)
Mass: ∗
∗
∗
∗
∂∂=
∂∂
tym ρ
(C. 3)
Boundary and initial conditions:
( ) ( ) ( ) 1,0,0, =∞== ∗∗∗∗∗∗ tTyyT ρ
( ) ( )11,0 * −Γ−=∂∂− ∗
∗
∗∗
sTtyTρ (C. 4)
Here, following non-dimensional parameters defined:
∞∗ = TTT /
∞∗ = ρρρ /
Lyy /=∗ ( )FkTL /∞∞= ρ
τ/tt =∗ ( ) pcL ∞∞∞∞ == ρλαατ /,/2
∞∗ = RTEE /
τAA =∗
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
217
∞∗ = ρρδ /c For char yield
Γ is the linearized heat loss term, containing the effects of convective and linearized
radiative heat losses from the surface.
( )( )( )11// 24 +++=Γ ∗∗∞∞ ss TTFTFhT σ
Here, ∗sT is an average surface temperature, chosen to match the heat losses to their actual
values over the appropriate temperature range.
Defined ( )∗∗∗ −= pTEAD /exp as the Damkohler number, thereby Equation (C. 2) can be
rewritten as:
( ) ( )
−−−−=
∂∂
∗
∗∗∗∗∗
∗
∗
1/expT
TTED
tpy
pyδρρ
For large values of ∗E , ∗
∗
∂∂
tρ
will be very small until ∗T is very close to the pyrolysis
temperature. Then ∗ρ =1 is used as approximate solution for the density for the stage 1 (inert
heating stage). Therefore, inert temperature can be obtained from equation (C. 1) and (C. 4).
+Γ×−
Γ+= Γ+Γ
*
**
*
**
22
11
*2*
t
yterfce
t
yerfcT ty
Inert (C. 5)
When *t reaches the pyrolysis time *pyt , the surface temperature ),0( ** tTInert equals *
pyT .
Assuming *pyt as a known parameter, for the condition )1(* Ox << and *t - *
pyt )1(O<< ,
Equation (C. 5) gives
⋅⋅⋅+−+−≅ )( ****pypy ttbayTT (C. 6)
and
⋅⋅⋅+−Γ++−≅∂∂
)]([ ****
*
pyttybayT
(C. 7)
where:
( )*
,0*
* *2
***
pyt
tty
Inert terfcey
Ta py
py
Γ=
∂∂−= Γ
==
(C. 8)
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
218
att
Tb
pytty
Inert
py
Γ−=
∂∂=
==*
,0*
* 1*** π
(C. 9)
and
)1(1
1),0( *** atTT pyInertpy −Γ
+== (C. 10)
Since pyrolysis always occurs on the surface first, *ρ =1, and most of changes of *ρ occur
near *pyT , the expressions in Equations (C. 5) and (C. 6) were used into Equation (C. 2). By
integrating over *t in the limit ∞→*E , following asymptotic evaluation result is achieved.
** /2**
**
/)1(
1 pyTE
pypy e
TbEA −−−≅ δρ
(C. 11)
For the second stage (initial pyrolysis), the basic equations (C. 1) to (C. 4) can be rewritten
as following by introducing a new item θ+= **InertTT :
(a)
∂∂+
∂∂
∂∂+
∂∂=
∂∂
**
*
*
*
*2*
2
*
1yy
Tyyt
Inert θρρ
θθ
(b)
−
+
−−−=
∂∂
1*exp)( *
*
*
***
*
*
θδρρ
Inert
py
py T
T
TE
Dt
(C. 12)
(c) ( ) **
*
*
**
*
)1(,0
ρθ
ρρθ Γ−
∂∂−−=
∂∂−
yT
ty
Inert
(d) *ρ ( *y ,0) = 1
θ ( *y ,0)= θ (∞, *t )=0
Similar calculation theme applied to the other stages, following expressions will be
obtained.
For the initial pyrolysis stage:
*ct = *
pyt +(b *E / *pyT 2)-1[ *s ( *δ )-lnDo] (C. 13)
where:
maxmax
** =′′=
mss
s=(b *E / *pyT 2)( *t - *
pyt )+lnDo-(a *E / *pyT 2) *y (C. 14)
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
219
Do=D/(b *E / *pyT 2)
Numerical solution shows that s depends only on *δ and the relation between them was
given in Figure C1. For *δ =0.3, *s ( *δ )=0.6.
Figure C1 Plot of *s ( *δ ) versus *δ
( ) ))(/(/2**
***
**2****
/1 pypypy ttTbETE
py
eTaE
Am −+−−≅′′ δ
(C. 15)
)1)(/( *max
* δ−=′′ abm (C. 16)
The thickness of the pyrolysis zone is given as following:
2**
*
/1.3
pyTaEy ≅∆
(C. 17)
For the thin-char stage, following equations can be obtained:
aterfce
yT
a
ct
tty
cc
c
c
)1
()( *
,0*
*
*2
***
δδ−+Γ=
∂∂−=
Γ
==
(C. 18)
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
220
bterfcet
tT
b
ct
c
tty
cc
c
c
)1
()(1 *
*
,0*
*2
***
δδ
π−+ΓΓ−=
∂∂=
Γ
==
∗
(C. 19)
( )( )
**
**
/2***
**2
*
*
max*
/1
1c
c
TEc
tts
c
c eTyT
AJ
Jm −
=∂∂−−
+=′′ δ
(C. 20)
where: 2
*
**
1.33.4
=
py
c
cc T
Taa
J (C. 21)
For the thick char stage, the reaction is assumed infinitely thin compared to the length scale
L. The surface temperature in this stage is assumed close to its maximum value as follows.
Γ+= /11*max,sT (C. 22)
Analysing mass balance across the reaction layer, below equation can be obtained.
**
*
*** )1()1( R
R vdtdy
m δδ −=−=′′ (C. 23)
where *Rv =d *
Ry /d *t is the velocity of the reaction layer.
Since the advance of a constant-temperature front into a solid is proportional to *t -0.5, its
rate of advance varies as *t -0.5. Then *m ′′ is also proportional to (1- *δ ) *t -0.5. The constant
of the proportionality, Ω, is solved by below equation.
For small Ω
−−=Ω
12 *
**max
*
c
c
TTTπδ
(C. 24)
Thus,
*
** )1(
tm
Ω−=′′ δ
(C. 25)
For a given set of input value, following parameters have been obtained from Atreya’ s
model, shown in Table C1. The values from numerical calculation are also listed. Excepting
the mass flux *m ′′ , difference between modelling and numerical results is within 20%.
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
221
The major calculation results for the mass flux by the given set of inputs were shown in
Figure C2.
Table C1 Atreya’s modelling results (input: *A =1011, *E =40, *δ =0.3, Γ =0.4)
Quantity Modelling results
Numerical value Notes
*pyT 1.5
*pyt 0.27
a 0.80 b 0.76
*ct 0.61 0.7
*maxm ′′ (initial pyrolysis) 0.66 0.825
*y∆ 0.22
ac 2.58 *ct =0.7 used
bc 2.16 *ct =0.7 used
*cT 1.76 1.97 *
ct =0.61 used *
cJ 0.74 *cT =1.97 used
*maxm ′′ (thin char) 0.73 0.825 *
cT =1.97 used *m ′′ */2.0 t */52.0 t
Figure C2 Mass fluxes from Atreya’s model
Appendix C Analysis and Computation of Atreya’ s Pyrolysis Model ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
222
Preparation calculation for one of the testing materials, mountain ash, was shown in Table
C2. The analytical calculation results for the mass loss fluxes were given in Chapter 3,
Section 3.5.4.
Table C2 Calculation results from applied Atreya’ s model Quantity Modelling Results Modified Value
*pyT 1.8
*pyt 0.51
a 0.65
b 0.6 *
ct 1.91 2.16 *
maxm ′′ (initial pyrolysis) 0.74 0.85 *m ′′ 1.62×107×e-29.7+9.64(t*-0.5) *y∆ 0.3
ac 2.08 bc 1.56
*cT 2.04 2.28
*cJ 0.53
*maxm ′′ (thin char) 0.88 1.02
*m ′′ */68.0 t */86.0 t
223
þÿ$ÿþÿjÿþÿ$ÿþÿjÿþÿ$ÿþÿjÿþÿ$ÿþÿjÿ
ABCB (1996), "Performance-Based Building Code of Australia", Australian Building Codes Board. Canberra, Australia.
Abu-Zaid, M. (1988), "Effect of water on piloted ignition of cellulosic materials". E. Lansing, MI, Michigan State University.
Agrawal, R. K. (1992), "Analysis of non-isothermal reaction kinetics: Part s. Complex reactions", Thermochimica Acta, 203: 111-125.
Alves, S. S. and J. L. Figueiredo (1989), "A model for pyrolysis of wet wood", Chemical Engineering Science, 44: 2861-2869.
Antal, M. J. and G. Varhegyi (1995), Ind. Eng. Chem. Res., 34(3): 703-717.
Arpaci, V. S. and P. S. Larsen (1984), "Convection Heat Transfer", Englewood Cliffs, NJ, Prentice-Hall Inc.
AS/NZS_3837 (1998), "Method of test for heat and smoke release rates for materials and products using an oxygen consumption calorimeter", Standards Australia & Standards New Zealand.
ASTM (1983), "Proposed method for room fire test of wall and ceiling materials and assemblies", ASTM Annual Book of Standards, 04.07: 958-978.
ASTM (1990), "Report to ASTM on Cone Calorimeter Inter-laboratory Trials", Task Group E5.21 T.G. 60.
ASTM_E176 "Standard Terminology of Fire Standards", Annual Book of ASTM Standard, ASTM. West Conshohocken, PA., Vol. 4.07.
ASTM_E1321-90 (1990), "Standard method for determining material ignition and flame spread properties", ASTM. West Conshohochen, PA.
ASTM_E1354-04 (2004), "Standard test method for heat and visible smoke release rates for materials and products using an oxygen consumption calorimeter", American Society for Testing and Materials. West Conshohocken, PA.
ASTM_E1582 (2000), "Standard Test Method for Rubber-Compositional Analysis by Thermogravimetry (TGA)".
Atreya, A. (1983), "Pyrolysis, ignition and fire spread on horizontal surface of wood". Cambridge, MA, Harvard University.
Atreya, A. (1984), "Pyrolysis, Ignition and Fire Spread on Horizontal Surface of Wood", NBS-GCR-83-449, National Bureau of Standard.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
224
Atreya, A., C. Carpentier and M. Harkleroad (1986), "Effect of sample orientation on piloted ignition and flame spread", Fire Safety Science - Proc. 1st Intl. Symp., Hemisphere, Washington.
Babrauskas, V. (1984), "Development of the Cone Calorimeter - A Bench-Scale Heat Release Rate Apparatus Based on Oxygen Consumption", Fire and Materials, 8: 81-95.
Babrauskas, V. (1992), "Generation of CO in Bench-Scale Fire Tests and the Prediction for Real-Scale Fires", Fire and Materials Int. Conf., Arlington, VA, Interscience Communications Ltd.
Babrauskas, V. (1996), "Fire Modelling Tools for FSE: Are They Good Enough?" Journal of Fire Protection Engineering, 8(2): 87-96.
Babrauskas, V. (2001), "Ignition of wood A review of the state of the art", Interflam 2001.
Babrauskas, V. (2002), "Extended Abstracts and Presentations from the Workshop on Fire Growth and Spread on Objects, March 4-6, 2002, National Institute of Standards and Technology", Workshop on Fire Growth and Spread on Objects, National Institute of Standards and Technology.
Babrauskas, V., J. Lawson, W. Walton and W. Twilley (1982), "Upholstered furniture heat release rates measured with a furniture calorimeter", NBSIR 82-2604, National Bureau of Standard.
Babrauskas, V. and R. D. Peacock (1992), "Heat Release Rate: the Single Most Important Variable in Fire Hazard", Fire Safety Journal, 18: 255-272.
Baker, R. R. (1978), "Kinetic parameters from the non-isothermal decomposition of a multi-component solid", Thermochimica Acta, 23: 201-212.
Bamford, C. H., J. Crank and D. H. Malan (1946), "On combustion of wood", Proc. Cambridge Phil. Soc., 42: 166-182.
Barooah, J. N. and V. D. Long (1976), Fuel, 55: 116.
Baum, H. R. (2002), "Extended Abstracts and Presentations from the Workshop on Fire Growth and Spread on Objects, March 4-6, 2002, National Institute of Standards and Technology", Workshop on Fire Growth and Spread on Objects, National Institute of Standards and Technology.
Beard, A. (1992), "Limitations of Computer Models", Fire Safety Journal, 18: 375-391.
Beard, A. N. (1997), "Fire models and design", Fire Safety Journal, 28: 117-138.
Beck, J. V. and K. J. Arnold (1977), "Parameter estimation in engineering and science", New York, John Wiley & Sons.
Beck, V. (1998), "CESARE-RISK: a tool for performance-based fire engineering design", 1998 Pacific Rim Conference and Second International Conference on Performance-Based Codes and Fire Safety Design Methods, Maui, Hawaii.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
225
Beyler, C. (2002), "Flammability Limits of Premixed and Diffusion Flames", The SFPF Handbook of Fire Protection Engineering: 2-173.
Beyler, C., S. P. Hunt and N. Iqbal (1997), "A computer model of upward flame spread on vertical surface", Fire Safety Science - Proc. 5th Interl. Symp., International Association of Fire Safety Science.
Beyler, C. L. and M. M. Hirschler (2002), "Thermal Decomposition of Polymer", The SFPF Handbook of Fire Protection Engineering: 1-110.
Bhargava, A., B. Z. Dlugogorski and E. M. Kennedy (2001), "Emissions from Fires of Wood Chips", Interflam 2001.
Bilbao, R., J. F. Mastral, M. E. Aldea, J. Ceamanos and J. A. Lana (2001), "Experimental and theoretical study of the ignition and smoldering of wood including convective effects", Combustion and Flame, 126: 1363-1372.
Blasi, C. D., S. Crescitelli, G. Russo and A. C. Fernandez-Pello (1989), 2nd Int. Symp. on Fire Safety Science, Hemisphere Pub. Co.
Bonnefoy, F., P. Gilot and G. Prado (1993), "A three-dimensional model for the determination of kinetic data from the pyrolysis of beech wood", Journal of Analytical and Applied Pyrolysis, 25: 387-394.
Bradbury, A. G. W., Y. Sakai and F. Shafizadeh (1979), "A kinetic model for pyrolysis of cellulose", Journal of Appl. Polym. Sci., 23: 3271-3280.
Brannigan, V. M. (1999), "Fire Scenarios or Scenario Fires? Can Fire Safety Science Provide the Critical Inputs for Performance Based Fire Safety Analysis?" Fire Safety Science, Proc. 6th Inter. Symp., Poitiers, France, International Association for Fire Safety Science.
Brauman, S. K. (1988), "Polymer degradation during combustion", Journal of Polymer Science, 26(B): 95-105.
Brown, M. E. (1988), "Introduction to Thermal Analysis: Techniques and Applications", London, New York, Chapman and Hall.
Bukowski, R. W. and V. Babrauskas (1990), "Developing rational, performance-based fire safety requirements in model building codes", Fire and Materials, 18: 173-191.
Burgess, M. J. and R. V. Wheeler (1911), J. Chem. Soc., 99: 2013.
Carrasco, F. (1993), "The evaluation of kinetic parameters from thermogravimetric data: comparison between established methods and the general analytical equation", Thermochimica Acta, 213: 115-134.
Carslaw, H. S. and J. C. Jaeger (1959), "Combustion of Heat in Solids", London, Oxford University press.
Chang, T. C., W. S. Shen, Y. S. Chiu and S. Y. Ho (1995), "Thermo-oxidative Degradation of Phosphorus-containing Polyurethane", Polymer Degradation and Stability, 49: 353-360.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
226
Chembukulam, S. K., A. S. Dandge, N. L. Kovllur, R. K. Seshagirl and R. Vaidyeswaran (1981), "Smokeless Fuel from Carbonized Sawdust", Ind. Eng. Chem. Prod. Res. Dev., 20: 714-719.
Chow, W. K. and R. Yin (2002), "Discussion on Two Plume Formulae with Computational Fluid Dynamics." Journal of Fire Sciences, 20(3).
Cleary, T. G., T. J. Ohlemiller and K. M. Villa (1992), "Influence of ignition source on the flaming fire hazard of upholstered furniture", NISTIR 4847, National Institute of Standard and Technology.
Cordero, T., F. Garcia and J. J. Rodriguez (1989), "A Kinetic Study of Holm Oak Wood Pyrolysis from Dynamic and Isothermal TG Experiments", Thermochimica Acta, 149: 225 - 237.
Cordero, T., J. M. Rodriguez_Maroto, F. Garcia and J. J. Rodriguez (1991), "Thermal decomposition of wood in oxidizing atmosphere", Thermochimica Acta, 191: 161-178.
Cordero, T., J. M. Rodriguez-Maroto, J. Rodriguez-Mirasol and J. J. Rodriguez (1990), "On the kinetic of thermal decomposition of wood and wood components", Thermochimica Acta, 164: 135-144.
Cordova, J. L., D. C. Walther, J. L. Torero and A. C. Fernandez-Pello (2001), Combustion Science and Technology, 164: 253.
Coward, H. F., C. W. Carpenter and W. Payman (1919), J. Chem. Soc., 115: 27.
Cox, G. (1994), "The challenge of fire modelling", Fire Safety Journal, 23: 123-132.
Cox, G., R. Chitty and S. Kumar (1989), "Fire modelling and the King's Cross fire investigation", Fire Safety Journal, 15: 103-106.
Cox, G. and S. Kumar (2002), "Modeling Enclosure Fires Using CFD", The SFPE Handbook. Section Three: 3-194.
Deepak, D. and D. Drysdale (1983), "Flammability of solids: an apparatus to measure the critical mass flux at firepoint", Fire Safety Journal, 5: 167-169.
DeHaan, J., "Our Changing World: Fires, Fuels, Investigations and Investigators. Part 1: Furnishings", http://www.interfire.org/features/ourchangingworld.asp, InterFire Online.
Delichatsios, M. A. (1999), "Ignition times for thermally thick and intermediate conditions in flat and cylindrical geometries", 6th International Symposium Fire Safety Science, IAFSS.
Delichatsios, M. A. (1999), "A new interpretation of data from LIFT (Lateral Ignition and Flame Transport) apparatus and modifications for creeping flame spread", Interflam '99.
Delichatsios, M. A. and K. Saito (1991), "Upward fire spread: Key flammability properties, similarity solutions and flammability indices", Fire Safety Science - Proceedings of the 3rd International Symposium.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
227
Dembsey, N. A., P. J. Pagni and R. B. Williamson (1995), "Compartment fire experiments: comparison with models", Fire Safety Journal, 25: 187-227.
DiBlasi, C. (1993), Combustion Science and Technology, 90: 315-340.
DiBlasi, C., B. Carmen and F. Enzo (1999), "Non-isothermal kinetics of wood degradation and combustion", Interflam'99.
Doyle, C. D. (1962), Applied Polymer Science, 6: 639.
Drysdale, D. (1999), "An Introduction to Fire Dynamics", Suffolk, John Wiley & Sons.
Drysdale, D. and H. E. Thomson (1989), "Flammability of plastics II: Critical mass flux at the firepoint", Fire Safety Journal, 14: 179-188.
Durbetaki, P., T. X. Phuoc, M. P. Mathur and J. M. Ekmann (1995), J. Numeric. Methods in Fluids, 20: 507-522.
Ellis, R. C. (1981), "Building Contents, the Real Fire Problem in the Room", Building Contents - The Real Fire Problem. Fall Conference, Philadelphia, PA.
Fernandez-Pello, A. C. (1977), Combustion Science and Technology, 17: 1.
Fernandez-Pello, A. C. (1995), "The solid phase", Combustion Fundamentals of Fire. G. Cox, Academic Press. Chapter 2.
Fernando, A. E. (2000), "Modelling of Fire Spread and the Growth of Fire in Buildings Using Computational Fluid Dynamics", Centre for Environmental Safety and Risk Engineering, Victoria University of Technology, Australia.
Flesischmann, C. M. and F. F. Chen (2001), "Radiant Ignition of Upholstered Furniture", The International Conference on Engineered Fire Protection Design, San Francisco, CA, SFPE & NIST.
Floyd, J. E., K. B. McGrattan, S. Hostikka and H. R. Baum (2003), "CFD fire simulation using mixture fraction combustion and finite volume radiative heat transfer", Journal of Fire Protection Engineering, 13: 11-36.
Flynn, J. H. and I. A. Wall (1966), J. Res. Nat. Bur. Stand., 70A: 487.
Forney, G. P. and K. B. McGrattan (2004), "User's Guide for Smokeview Version 4 - A Tool for Visualizing Fire Dynamics Simulation Data", NIST Special Pub. 1017, NIST.
Fredlund, B. (1988), "A Model for Heat and Mass Transfer in Timber Structures during Fire, A Theoretical, Numerical and Experimental Study", LUTVDG/(TVBB-1003), Dept. of Fire Engineering, Lund University, Sweden.
Friday, P. and F. W. Mowrer (2001), "Comparison of FDS model predictions with FM/SNL fire test data", NIST GCR 01-810, National Institute of Standards and Technology.
Friedman, R. (1992), "An International Survey of Computer Models for Fire and Smoke", Journal of Fire Protection Engineering, 4(3): 81-92.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
228
FTT (1998), "User's Guide for the Cone Calorimeter", CONE WPD, Fire Testing Technology Limited.
Gijsman, P., J. Hennekens and J. Vincent (1993), "The mechanism of the low-temperature oxidation of polypropylene", Polymer Degradation and Stability, 42: 95-105.
Goos, A. (1952), "The thermal decomposition of wood", Wood Chemistry, 2.
Grexa, O., E. Horvathova and A. Osvald (1997), "Cone calorimeter studies of wood species", ISFST '97.
Grexa, O., M. Janssens, R. White and M. Dietenberger (1996), "Fundamental Thermophysical Properties of Materials Derived from the Cone Calorimeter Measurements", 3rd International Scientific Conference Wood & Fire Safety, The High Tatras Hotel Patria, Slovakia.
Gullet, B. K. and P. Smith (1987), Combustion and Flame, 67: 143.
Haessler, W. M. (1974), "The tetrahedron of fire", The Extinguishment of Fire. M. A. Quincy, National Fire Protection Association.
Hakkarainen, T. and M. A. Kokkala (2001), "Application of a one-dimensional thermal flame spread model on predicting the rate of heat release in the SBI test", Fire and Materials, 25: 61-70.
Hall, G. S., R. G. Saunders, R. T. Allcorn, P. E. Jackman, M. W. Hickey and R. Fitt (1972), "Fire Performance of Timber - A Literature Survey", Timber Research and Development Association.
Hamins, A., A. Maranghides, K. B. McGrattan, E. Johnsson, T. J. Ohlemiller, M. Donnelly, J. Yang, G. Mulholland, K. Parasad, S. Kukuck, R. Anleitner and T. McAllister (2004), "Report on Experiments to Validate Fire Dynamic and Thermal-Structural Models for Use in the World Trade Center Investigation", NIST Special Publication in preparation, NIST.
Harada, T. (2001), "Time to ignition, heat release rate and fir endurance time of wood in cone calorimeter test", Fire and Materials, 25: 161-167.
He, Y. (1996), "Further development and modification of the NRCC fire growth model", CESARE Report IR96-002, CESARE, Victoria University of Technology, Australia.
He, Y., Y. Jiang, M. Luo, V. Dowling, P. Bowditch and J. Leonard (1999), "Heat release and product yields of polyurethane foams under fabric cover (Abstract)", Interflam '99.
He, Y., I. Morre, M. Luo and V. Beck (1998), "Doorway Calorimetry for Heat Release Rate Measurement in Building Fires", Combustion Science and Technology, 132: 365-389.
Heskestad, G. (2002), "Plumes, Flame Height and Air Entrainment", The SFPE Handbook of Fire Protection Engineering, 3rd edition, National Fire Protection Association.
Hietaniemi, J., S. Hostikka and J. Vaari (2004), "FDS Simulation of Fire Spread - Comparison of Model Results with Experimental Data", Technical Report VTT Working Paper 4, VTT Building and Transport.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
229
Hostikka, S. (2002), "Kinetic Modelling of Wood Pyrolysis in FDS2.2", Workshop on Fire Growth and Spread on Objects, NIST.
Hostikka, S. (2004), Personal communication, Email.
Huggett, C. (1980), "Estimation of the rate of heat release by means of oxygen consumption measurements", Fire and Materials, 4: 61-65.
Ierardi, J. A. and J. R. Barnett (2003), "A quantitative method for calibrating CFD model calculations", CIB-CTBUH International conference on Tall Buildings, International Council for Research and Innovation in Building and Construction (CIB).
ISO_5660-1 (2002), "Fire tests - Reaction to Fire - Part 1: Rate of heat release from building products (Cone Calorimeter method)", ISO, Technical Committee ISO/TC 92.
ISO_9705 (1993), "Full-scale room fire test for surface products", International Organization for Standardization. Geneva.
Janssens, M. (1991), "Fundamental thermophysical characteristics of wood and their role in enclosure fire growth". Washington DC, National Forest Products Association.
Janssens, M. (1992), "Determining flame spread properties from cone calorimeter measurements", Heat release in fires. V. Babrauskas and S. J. Grayson. London and New York, Elsevier Applied Science. 1: 265-281.
Janssens, M. (1994), "Thermo-physical properties for wood pyrolysis models", Pacific Timber Engineering Conference, Gold Coast, Australia.
Janssens, M. (2002), "Calorimetry", The SFPE Handbook of Fire Protection Engineering, 3rd edition, SFPE. Section 3/Chapter 2.
Janssens, M. L. (2004), "Modeling of the thermal degradation of structural wood members exposed to fires", Fire and Materials, 28: 199-207.
Janssens, M. L., S. E. Dillon and S. Allwein (2001), "Characterizing the thermal environment of the cone calorimeter for analyzing ignition data of materials", Interflam 2001.
Jones, W. and G. Forney (1990), "A programmer's reference manual for CFAST, the unified model of fire growth and smoke transport", NIST Technical Note 1283, National Institute of Standards and technology.
Jones, W. W. (1983), "A review of compartment fire models", NBSIR 83-2684, National Bureau of Standards.
Jones, W. W. (2000), "Modeling Fires - the next generation of tools", 15th Meeting of the UJNR Panel on Fire Research and Safety, NIST.
Jones, W. W. (2001), "State of the art in zone modeling of fires", 9th International fire Protection Seminar, Munich, Germany.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
230
Jonsson, R. and O. Pettersson (1985), "Timber Structures and Fire - A Review of the Existing State of Knowledge and Research Requirements", Swedish Council for Building Research.
Kanury, A. M. (1994), Combustion Science and Technology, 97: 469-491.
Kanury, A. M. (2002), "Flaming Ignition of Solid Fuels", The SFPF Handbook of Fire Protection Engineering, 3rd edition: 2.229-2.245.
Kanury, M. A. (1995), "Flaming Ignition of Solid Fuels", The SFPF Handbook of Fire Protection Engineering: pp 2-190 to 2-204.
Kanury, M. A. and P. L. Blackshear (1970), "Some considerations pertaining to the problem of wood-burning", Combustion and Technology, 1: 339-356.
Kashiwagi, T. (1974), Combustion Science and Technology, 8: 225.
Kashiwagi, T., T. Hirata and J. E. Brown (1985), "Thermal and oxidative degradation of Poly(methyl methacrylate), molecular weight", Macromolecules, 18: 131-138.
Kashiwagi, T. and H. Nambu (1992), "Globe kinetic constants for thermal oxidative degradation of a cellulosic paper", Combustion and Flame, 88: 345-368.
Kashiwagi, T. and A. Omori (1988), "Effects of thermal stability and melt viscosity of thermoplastics on piloted ignition", 22nd Symp. Int. Combustion.
Kashiwagi, T., A. Omori and H. Nanbu (1990), Combustion and Flame, 81: 188-201.
Kissinger, H. E. (1957), Analytical Chemistry, 29: 1702.
Klose, W., S. Damm and W. Wiest (2000), "Pyrolysis and activation of different woods - thermal analysis (TG/EGA) and formal kinetics (Oral presentation)", 4th Int. Symp. Catalytic and Thermochemical Conversions of Natural Organic Polymers, Krasnoyarsk.
Kung, H. C. (1972), "A mathematical model of wood pyrolysis", Combustion and Flames, 18: 185-195.
Launder, B. E. and D. B. Spalding (1974), "The Numerical Computation of Turbulent Flows", Computer Methods in Applied Mechanics and Engineering, 3: 269-289.
Lawson, J. R. and T. A. Pinder (2000), "Estimates of Thermal Conductivity for Materials Used in Fire Fighters' Protective Clothing", NISTIR 6512, National Institute of Standards and Technology.
Lee, C., R. Chaiken and J. M. Singer (1976), "Soot and hydrocarbon formation in a turbulent diffusion flame", 16th Symp. (Int.) Comb.
Lee, C. K., R. F. Chaiken and J. M. Singer (1977), "Charring in Pyrolysis of Wood in Fires by Laser Simulation", 6th Symp. (Int.) Comb., The Combustion Institute.
Li, Y. and D. Drysdale (1992), "Measurement of the ignition temperature of wood", Fire Science and Technology - Proc. 1st Asian Conf. Intl., Beijing, Academic Publishers.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
231
Liu, N. and W. Fan (1998), "Modelling the Thermal Decomposition of Wood and Leaves Under a Nitrogen Atmosphere", Fire and Materials, 22(3): 103-108.
Liu, N., B. Wang and W. Fan (2003), "Kinetic Compensation Effect in the Thermal Decomposition of Biomass in Air Atmosphere", Fire Safety Science, 7th IAFSS, Worcester, MA, IAFSS.
Liu, N. A. (2000), "Kinetics of thermal decomposition for biomass and relevant kinetic analysis method", Uni. Sci. Tech. China.
Lucon, O. S. (1997), "Preliminary Environmental Performance Analysis of a Small-scale Combined Heat and Power System Attached to a Wood Chips Gasification Unit", Department of Chemical and Process Engineering, University of Newcastle.
Luo, M. and V. R. Beck (1996), "Effects of Air-handling System on Fire Spread in a Full-scale Building", 6th Australian Heat and Mass Transfer Conference.
Luo, M. and Y. He (1997), "Kinetics of pyrolysis and combustion of polyurethane foams under fire conditions", Fifth Inter. Symp. Fire Safety Science, Tsukuba, Japan.
Lyon, R. E. (2004), "Fire Safety", Highlights 2004, Federal Aviation Administration, U.S. Department of Transportation.
Lyon, R. E. and M. L. Janssens (2005), "Polymer Flammability", DOT/FAA/AR-05/14, Federal Aviation Administration, U.S. Department of Transportation.
Martin, S. (1965), "Diffusion-controlled Ignition of Cellulosic Materials", 10th Symp. (Inter.) on Comb., The Combustion Institute.
Martin, S. (1975), "Characterization of the Stanford Research Institute Large-Scale Heat-Release Rate Calorimeter", NBS-GCR 76-54, National Bureau of Standards.
McGrattan, K. and G. Forney (2005), "Fire Dynamics Simulator (Version 4) User's Guide", NIST Special Publication 1019, National Institute of Standard and Technology.
McGrattan, K. B. (2004), Personal communication, Conversation.
McGrattan, K. B. (2004), "Fire Dynamics Simulator (Version 4) Technical Reference Guide", NIST Special Publication 1018, Fire Research Division, Building and Fire Research Laboratory, NIST.
McGrattan, K. B., H. R. Baum and R. G. Rehm (1998), "Large Eddy Simulations of Smoke Movement", Fire Safety Journal, 30: 161-178.
McGrattan, K. B., H. R. Baum, R. G. Rehm, A. Hamins and G. P. Forney (2000), "Fire Dynamics Simulator, Technical Reference Guide", Technical Report NISTIR 6467, National Institute of Standards and Technology.
Messerschmidt, B., P. V. Hees and U. Wickstrom (1999), "Prediction of SBI (Single Burning Item) test results by means of cone calorimeter test results", Interflam '99.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
232
Mierlo, R. v. (1997), "SBI Round-Robin: general information", TNO Building and Construction Research.
Mierlo, R. v. (1998), "Development of the SBI test method (Volume 1)", TNO-report 98-CVB-R1067, TNO Building and Construction Research.
Mikkola, E. (1991), Fire Safety Science - 3rd Int. Symp.
Mitler, H. E. (1985), "The HARVARD fire model", Journal of Fire Safety, 9: 7-16.
Mitler, H. E. and J. A. Rockett (1987), "Users guide to FIRST", NBSIR 87-3595, National Bureau of Standards.
Moghtaderi, B. (2001), "Pyrolysis of Char Forming Solid Fuels: A Critical Review of the Mathematical Modelling Techniques", 5th AOSFST, Newcastle, Australia.
Moghtaderi, B., B. Z. Dlugogorski, E. M. Kennedy and D. F. Fletcher (1998), "Effects of the structural properties of solid fuels on their re-ignition characteristics", Fire and Materials, 22: 155-165.
Moghtaderi, B., V. Novozhilov, D. F. Fletcher and J. H. Kent (1997), "An integral model for the transient pyrolysis of solid materials", Fire and Materials, 21(1): 7-16.
Moghtaderi, B., V. Novozhilov, D. F. Fletcher and J. H. Kent (1997), "A new correlation for bench-scale piloted ignition data of wood", Journal of Fire Safety, 29: 41-59.
Mukaro, R., M. Gasseller, C. Kufazvinei, L. Olumekor and B. M. Taele (2003), "Microcontroller-based Multi-sensor apparatus for temperature control and thermal conductivity measurement", Measurement Science and Technology, 14: N45-N49.
Niioka, T. (1981), "Gas-phase ignition of a solid fuel in a hot stagnation-point flow", 18th Symp. Int. Combust.
Nurbakhsh, S. (1989), "Thermal decomposition of charring materials". E. Lansing, MI, Michigan University.
Ohlemiller, T. J. and J. R. Shields (1995), "Behavior of mock-ups in the California Technical Bulletin 133 test protocol: fabric and barrier effects", NISTIR 5653, National Institute of Standard and Technology.
Ohlemiller, T. J. and M. Summerfield (1971), 13th Symp. Int. Combust.
Omrane, A., F. Ossler, M. Alden, U. Goransson and G. Holmstedt (2003), "Surface temperature measurement of flame spread using thermographic phosphors", Fire Safety Science - Proceedings of the 7th International Symposium.
Ozawa, T. (1970), Journal of Thermal Analysis, 2: 301.
Pape, P., T. E. Waterman and T. V. Eichler (1981), "Development of a fire in a room from ignition to full room involvement - fires", NBS-GCR-81-301, National Bureau of Standards.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
233
Parker, W. and M. Long (1972), "Development of a Heat Release Rate Calorimeter at NBS", ASTM.
Parker, W. J. (1985), "Development of a model for the heat release rate of wood - A status report", NBSIR 85-3163, National Bureau of Standards.
Parker, W. J. (1985), "Prediction of heat release rate of wood", Fire Safety Science - 1st Int. Symp.
Parker, W. J. (1988), "Determination of input data for a model of the heat release rate of wood", Mathematical modeling of fires. J. R. Mehaffey. Philadelphia, PA, ASTM: 105-115.
Parker, W. J. (1989), "Prediction of the heat release rate of Douglas Fir", 2nd Int. Symp. Fire Safety Science.
Parker, W. J. (1992), "Prediction of the heat release rate from basic measurements", Heat Release in Fires. V. Babrauskas and S. J. Grayson. London and New York, Elsevier Applied Science. 2: 333-356.
Parker, W. J. and S. L. LeVan (1989), "Kinetic properties of the components of Douglas-fir and the heat of combustion of their volatile pyrolysis products", Wood and Fiber Science, 21(3): 289-305.
Peacock, R. D., G. P. Forney, P. Reneke, R. Portier and W. W. Jones (1993), "CFAST, the Consolidated Model of Fire Growth and Smoke Transport", NIST Tech. Note 1299, National Institute of Standards and Technology.
Peacock, R. D., W. W. Jones, R. W. Bukowski and C. L. Forney (1991), "Technical Reference Guide for the HAZARD I Fire Hazard Assessment Method", NIST Hbbk 146, National Institute of Standards and Technology.
Perkin_Elemer (1991), "TGA 7 Thermogravimetric Analyzer Operator's Manual", Perkin Elemer Corporation.
prEN_13823 (2002), "Reaction to fire tests for building products - Building products excluding floorings exposed to the thermal attack by a single burning item", European Committee for Standardization. Brussels.
Quintiere, J. G. (1977), "Growth of fires in building compartments", ASTM STP 614, American Society for Testing and Materials.
Quintiere, J. G. (1981), "A simplified theory for generalized results from a radiant panel rate of flame spread apparatus (LIFT)", Fire and Materials, 5(2): 52-60.
Quintiere, J. G. (1988), "The application of flame spread theory to predict material performance", Journal of Research of the NBS, 93(1): 61-70.
Quintiere, J. G. (1993), "A simulation model for fire growth on materials subject to a room-corner test", Fire Safety Journal, 20: 313-339.
Quintiere, J. G. (1995), "Surface Flame Spread", SFPE Handbook. Section 2, Chapter 14: 2-205.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
234
Quintiere, J. G. (2002), "Compartment fire modeling", The SFPE Handbook of Fire Protection Engineering, 3rd edition. Section 3, Chapter 5: 3-125.
Quintiere, J. G. and M. Harkleroad (1984), "New Concepts for Measuring Flame Spread Properties", NBSIR 84-2943, US. Department of Commerce.
Quintiere, J. G. and B. Rhodes (1994), "Fire growth models for materials", NIST-GCR-94-647, National Institute of Standards and Technology.
Raithby, G. D. and E. H. Chui (1990), "A Finite-Volume Method for Predicting Radiant Heat Transfer in Enclosures with Participating Media", Journal of Heat Transfer, 112(2): 415-423.
Ramiah, M. V. (1970), "Thermogravimetric and differential thermal analysis of cellulose, hemicellulose, and lignin", Journal of Appl. Polym. Sci., 14: 1323-1337.
Rasbash, D. J., D. D. Drysdale and D. Deepak (1986), "Critical heat and mass transfer st pilot ignition of a material", Fire Safety Journal, 10: 1-10.
Richardson, J. K. (1994), "Moving towards performance-based codes", NFPA Journal, 88(May/June): 70-78.
Ritchie, S. J., K. D. Steckler, A. Hamins, T. G. Cleary and J. C. Kashiwagi (1997), "Effect of Sample Size on the Heat Release Rate of Charring Materials", Fire Safety Science. Proc. 5th Inter. Symp., Melbourne, Australia.
Roberts, A. F. (1970), "A Review of Kinetics Data for the Pyrolysis of Wood and Related Substances", Combustion and Flame, 14: 261-272.
Sauer, F. (1956), "The charring of wood during exposure to thermal radiation: correlation analysis for semi-infinite solids", Intrim, T. AFSWP-868, Forest Service.
Schaffer, E. L. (1967), "Charring rate of selected woods - transverse to grain", Research Paper FPL 69, U.S. Department of Agriculture, Forest Products Laboratory.
Schneider, H. A. (1992), "Thermogravimetric kinetics of polymer degradation: science or fiction?" Polymer Engineering and Science, 32(17): 1309-1315.
Schultz, T. P., G. D. McGinnis and D. Nicholas, Eds. (1989), "Effect of crystallinity and additives on the thermal degradation of cellulose", Fire and Polymers Hazard Identification and Prevention, American Chemical Society.
Seungdo, K. and J. K. Park (1995), "Characterization of thermal reaction by peak temperature and height of DTG curves", Thermochimica Acta, 264: 137-156.
SFPE (2002), "SFPE Handbook of Fire Protection Engineering".
Simms, D. L. (1963), "On the pilot ignition of wood by radiation", Combustion and Flame, 7: 253.
Smith, E. (1972), "Heat Release Rate of Building Materials", ASTM.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
235
Snell, J. E., V. Babrauskas and A. J. Fowell (1993), "Elements of a framework for fire safety engineering", Interflam '93, London, Interscience Communications Ltd.
Spearpoint, M. J. and J. G. Quintiere (2000), "Predicting the burning of wood using an integral model", Combustion and Flame, 123: 308-324.
Spearpoint, M. J. and J. G. Quintiere (2001), "Predicting the Piloted Ignition of Wood in the Cone Calorimeter Using an Integral Model - Effect of Species, Grain Orientation and Heat Flux", Fire Safety Journal, 36: pp 391 - 415.
Staggs, J. E. J. (1999), "Modelling thermal degradation of polymers using single-step first-order kinetics", Fire Safety Journal, 32: 17-34.
Staggs, J. E. J. and R. H. Whiteley (1999), "Modelling the combustion of solid-phase fuels in cone calorimeter experiments", Fire and Materials, 23: 63-69.
Stanmore, B. R. (1991), "Use of the Differential Thermal Gravimetric Analyser for Carbon/Oxygen Reactivity Measurement", Fuel, 70: 1485-1487.
Steckler, K. D., A. Hamins and T. Kashiwagi (1994), "A model for the burning of a horizontal slab of wood: status report", NISTIR 5499, NIST.
Stroup, D. W. (2002), "Using Field Modeling to Simulate Enclosure Fires", Boston, MA, Society of Fire Protection Engineers.
Tanaka, T. (1983), "A model of Multicompartment Fire Spread", NBSIR 83-2718, National Beaure of Standards.
Tang, W. K. (1972), "Forest Products", Differential Thermal Analysis. R. C. Mackenzie, Academic Press, New York. 2: 523-553.
Tewarson, A. (1982), "Experimental evaluation of flammability parameters of polymeric materials", Flame retardant polymeric materials. E. M. Pearce. New York, London, Plenum Press. 3: 97-153.
Tewarson, A., W. Chin and R. Shuford (1999), "An exploratory study on a high-energy flux (HEF) calorimeter to characterize flammability of advanced engineered polymers: phase 1 - ignition and mass loss rate", ARL-TR-2102.
Thompson, N. and E. Cousins (1959), "The FM Construction Materials Calorimeter", NFPA Quarterly, 52: 186-192.
Thomson, H. E. and D. D. Drysdale (1989), "Flammability of plastics II Critical mass flux at the firepoint", Fire Safety Journal, 14(3): 179-188.
Thornton, W. (1917), "The Relation of Oxygen to the Heat of Combustion of Organic Compounds", Philosophical Magazine and J. of Science, 33.
T'ien, J. S. (2002), "Detailed modeling of flame spread processes over solid: progress and prospect", Workshop on Fire Growth and Spread on Objects, March 4-6, 2002, National Institute of Standards and Technology.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
236
Tran, H. C. (1992), "Experimental data on wood materials", Heat release in fires. V. Babrauskas and S. J. Grayson. New York, Elsevier Applied Science. Chapter 11.
Tsai, T., M. Li, I. Shih, R. Jih and S. Wong (2001), "Experimental and numerical study of autoignition and pilot ignition of PMMA plates in a cone calorimeter", Combustion and Flame, 124: 466-480.
Tzeng, L., A. Atreya and I. S. Wichman (1990), "A one-dimensional model of piloted ignition", Combustion and Flame, 80: 94-107.
Urbas, J. and W. J. Parker (1993), "Surface temperature measurements on burning wood specimens in the cone calorimeter and the effect of grain orientation", Fire and Materials, 17: 205-208.
Urbas, J. and W. J. Parker (2005), "Surface Temperature Measurement in a Fire Environment Using an Infrared Pyrometer", Fire Safety Science - Proceedings of the 8th International Symposium.
Urbas, J., W. J. Parker and G. E. Luebbers (2004), "Surface temperature measurements on burning materials using an infrared pyrometer: accounting for emissivity and reflection of external radiation", Fire and Materials, 28: 33-53.
Vervisch, L. and T. Poinsot (1998), "Direct numerical simulation of non-premixed turbulent flames", Ann. Rev. Fluid Mech., 30: 655.
Vovelle, C., R. Akrich and J. L. Delfau (1984), 20th Smpy. (Int.) Comb.
VTT (1997), "Development of the Single Burning Item test - results of the SBI Round Robin tests", VTT Building Technology, Fire Technology.
Vyazovkin, S. and C. A. Wight (1998), "Isothermal and non-isothermal kinetics of thermally stimulated reactions of solids", International Reviews in Physical Chemistry, 17(3): 407-433.
Wagoner, C. L. and A. F. Duzy (1976), "Burning Profiles of Solid Fuels", ASME Paper 67-WA/FU-4, ASME.
Walther, D. C., R. A. Anthenien and A. C. Fernandez-Pello (2000), "Smolder ignition of polyurethane foam: effect of oxygen concentration", Fire Safety Journal(34): 343-359.
Walton, W. D. (1985), "ASET-B A room fire program for personal computers", NBSIR 85-3144, National Bureau of Standards.
Wang, G. G. and J. T. Yang (1992), "Convective thermal ignition of an individual polymethylmethacrylate particle", Journal of Thermophysics and Heat Transfer, 6(2): 326-332.
Watt, S. D., J. E. J. Staggs, A. C. McIntosh and J. Brindley (2001), "A theoretical explanation of the influence of char formation on the ignition of polymers", Fire Safety Journal, 36: 421-436.
References ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
237
White, R. and E. Nordheim (1992), "Charring rate of wood for ASTM E119 exposure", Fire Technology, 28: 5-30.
Wichman, I. S. (1986), Combustion and Flame, 63: 217-229.
Williams, F. A. (1976), "Mechanisms of fire spread", 16th Symp. Int. Combustion.
Williams, F. A. (1985), "Combustion Theory", The Benjamin/Cummings Co. Inc.
WPI, "Cone Calorimeter Data", http://www.wpi.edu/Academics/Depts/Fire/Lab/Cone/Data, Accessed May 1, 2005, Fire Protection Engineering & Center for Firesafety Studies.
Xue, H., J. C. Ho and Y. M. Cheng (2001), "Comparison of different combustion models in enclosure fire simulation", Fire Safety Journal, 36: 37-54.
Yang, L., X. Chen, X. Zhou and W. Fan (2003), "The pyrolysis and ignition of charring materials under an external heat flux", Combustion and Flame, 133: 407-413.
Yao, Y. B., C. Ryu, D. Poole, P. Gilbert, A. Khor, V. Sharifi and J. Swithenbank, "Mathematical Modelling of Biomass Pyrolysis and Gasification in Lab-scale Packed-bed Systems", Mathmatical Modelling Workshop, SUWIC, Sheffield University.
Yuen, R., R. Casey, G. d. V. Davis, E. Leonardi, G. H. Yeoh, V. Chandrasekaran and S. J. Grubits (1997), "Three-dimensional mathematical model for the pyrolysis of wet wood", Fire Safety Science - Proc. 5th Interl. Symp., International Association of Fire Safety Science.
Zhou, X., K. H. Luo and J. J. R. Williams (2000), "Dynamic behaviour in reacting plumes", 28th Symposium (International) on Combustion, Pittsburgh, Combustion Institute.
Zhou, Y. Y., D. C. Walther and A. C. Fernandez-Pello (2002), "Numerical Analysis of Piloted Ignition of Polymeric Materials", Combustion and Flame, 131: 147-158.